diff.con commit diff: tweak a _copy_ of diff_options with word-diff (6440d34)
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13#include "utf8.h"
  14#include "userdiff.h"
  15#include "sigchain.h"
  16#include "submodule.h"
  17#include "ll-merge.h"
  18
  19#ifdef NO_FAST_WORKING_DIRECTORY
  20#define FAST_WORKING_DIRECTORY 0
  21#else
  22#define FAST_WORKING_DIRECTORY 1
  23#endif
  24
  25static int diff_detect_rename_default;
  26static int diff_rename_limit_default = 400;
  27static int diff_suppress_blank_empty;
  28int diff_use_color_default = -1;
  29static const char *diff_word_regex_cfg;
  30static const char *external_diff_cmd_cfg;
  31int diff_auto_refresh_index = 1;
  32static int diff_mnemonic_prefix;
  33static int diff_no_prefix;
  34static int diff_dirstat_permille_default = 30;
  35static struct diff_options default_diff_options;
  36
  37static char diff_colors[][COLOR_MAXLEN] = {
  38        GIT_COLOR_RESET,
  39        GIT_COLOR_NORMAL,       /* PLAIN */
  40        GIT_COLOR_BOLD,         /* METAINFO */
  41        GIT_COLOR_CYAN,         /* FRAGINFO */
  42        GIT_COLOR_RED,          /* OLD */
  43        GIT_COLOR_GREEN,        /* NEW */
  44        GIT_COLOR_YELLOW,       /* COMMIT */
  45        GIT_COLOR_BG_RED,       /* WHITESPACE */
  46        GIT_COLOR_NORMAL,       /* FUNCINFO */
  47};
  48
  49static int parse_diff_color_slot(const char *var, int ofs)
  50{
  51        if (!strcasecmp(var+ofs, "plain"))
  52                return DIFF_PLAIN;
  53        if (!strcasecmp(var+ofs, "meta"))
  54                return DIFF_METAINFO;
  55        if (!strcasecmp(var+ofs, "frag"))
  56                return DIFF_FRAGINFO;
  57        if (!strcasecmp(var+ofs, "old"))
  58                return DIFF_FILE_OLD;
  59        if (!strcasecmp(var+ofs, "new"))
  60                return DIFF_FILE_NEW;
  61        if (!strcasecmp(var+ofs, "commit"))
  62                return DIFF_COMMIT;
  63        if (!strcasecmp(var+ofs, "whitespace"))
  64                return DIFF_WHITESPACE;
  65        if (!strcasecmp(var+ofs, "func"))
  66                return DIFF_FUNCINFO;
  67        return -1;
  68}
  69
  70static int parse_dirstat_params(struct diff_options *options, const char *params,
  71                                struct strbuf *errmsg)
  72{
  73        const char *p = params;
  74        int p_len, ret = 0;
  75
  76        while (*p) {
  77                p_len = strchrnul(p, ',') - p;
  78                if (!memcmp(p, "changes", p_len)) {
  79                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  80                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  81                } else if (!memcmp(p, "lines", p_len)) {
  82                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  83                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  84                } else if (!memcmp(p, "files", p_len)) {
  85                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  86                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  87                } else if (!memcmp(p, "noncumulative", p_len)) {
  88                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  89                } else if (!memcmp(p, "cumulative", p_len)) {
  90                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
  91                } else if (isdigit(*p)) {
  92                        char *end;
  93                        int permille = strtoul(p, &end, 10) * 10;
  94                        if (*end == '.' && isdigit(*++end)) {
  95                                /* only use first digit */
  96                                permille += *end - '0';
  97                                /* .. and ignore any further digits */
  98                                while (isdigit(*++end))
  99                                        ; /* nothing */
 100                        }
 101                        if (end - p == p_len)
 102                                options->dirstat_permille = permille;
 103                        else {
 104                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%.*s'\n"),
 105                                            p_len, p);
 106                                ret++;
 107                        }
 108                } else {
 109                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%.*s'\n"),
 110                                    p_len, p);
 111                        ret++;
 112                }
 113
 114                p += p_len;
 115
 116                if (*p)
 117                        p++; /* more parameters, swallow separator */
 118        }
 119        return ret;
 120}
 121
 122static int git_config_rename(const char *var, const char *value)
 123{
 124        if (!value)
 125                return DIFF_DETECT_RENAME;
 126        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 127                return  DIFF_DETECT_COPY;
 128        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 129}
 130
 131/*
 132 * These are to give UI layer defaults.
 133 * The core-level commands such as git-diff-files should
 134 * never be affected by the setting of diff.renames
 135 * the user happens to have in the configuration file.
 136 */
 137int git_diff_ui_config(const char *var, const char *value, void *cb)
 138{
 139        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 140                diff_use_color_default = git_config_colorbool(var, value);
 141                return 0;
 142        }
 143        if (!strcmp(var, "diff.renames")) {
 144                diff_detect_rename_default = git_config_rename(var, value);
 145                return 0;
 146        }
 147        if (!strcmp(var, "diff.autorefreshindex")) {
 148                diff_auto_refresh_index = git_config_bool(var, value);
 149                return 0;
 150        }
 151        if (!strcmp(var, "diff.mnemonicprefix")) {
 152                diff_mnemonic_prefix = git_config_bool(var, value);
 153                return 0;
 154        }
 155        if (!strcmp(var, "diff.noprefix")) {
 156                diff_no_prefix = git_config_bool(var, value);
 157                return 0;
 158        }
 159        if (!strcmp(var, "diff.external"))
 160                return git_config_string(&external_diff_cmd_cfg, var, value);
 161        if (!strcmp(var, "diff.wordregex"))
 162                return git_config_string(&diff_word_regex_cfg, var, value);
 163
 164        if (!strcmp(var, "diff.ignoresubmodules"))
 165                handle_ignore_submodules_arg(&default_diff_options, value);
 166
 167        if (git_color_config(var, value, cb) < 0)
 168                return -1;
 169
 170        return git_diff_basic_config(var, value, cb);
 171}
 172
 173int git_diff_basic_config(const char *var, const char *value, void *cb)
 174{
 175        if (!strcmp(var, "diff.renamelimit")) {
 176                diff_rename_limit_default = git_config_int(var, value);
 177                return 0;
 178        }
 179
 180        switch (userdiff_config(var, value)) {
 181                case 0: break;
 182                case -1: return -1;
 183                default: return 0;
 184        }
 185
 186        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 187                int slot = parse_diff_color_slot(var, 11);
 188                if (slot < 0)
 189                        return 0;
 190                if (!value)
 191                        return config_error_nonbool(var);
 192                color_parse(value, var, diff_colors[slot]);
 193                return 0;
 194        }
 195
 196        /* like GNU diff's --suppress-blank-empty option  */
 197        if (!strcmp(var, "diff.suppressblankempty") ||
 198                        /* for backwards compatibility */
 199                        !strcmp(var, "diff.suppress-blank-empty")) {
 200                diff_suppress_blank_empty = git_config_bool(var, value);
 201                return 0;
 202        }
 203
 204        if (!strcmp(var, "diff.dirstat")) {
 205                struct strbuf errmsg = STRBUF_INIT;
 206                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 207                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 208                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 209                                errmsg.buf);
 210                strbuf_release(&errmsg);
 211                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 212                return 0;
 213        }
 214
 215        if (!prefixcmp(var, "submodule."))
 216                return parse_submodule_config_option(var, value);
 217
 218        return git_default_config(var, value, cb);
 219}
 220
 221static char *quote_two(const char *one, const char *two)
 222{
 223        int need_one = quote_c_style(one, NULL, NULL, 1);
 224        int need_two = quote_c_style(two, NULL, NULL, 1);
 225        struct strbuf res = STRBUF_INIT;
 226
 227        if (need_one + need_two) {
 228                strbuf_addch(&res, '"');
 229                quote_c_style(one, &res, NULL, 1);
 230                quote_c_style(two, &res, NULL, 1);
 231                strbuf_addch(&res, '"');
 232        } else {
 233                strbuf_addstr(&res, one);
 234                strbuf_addstr(&res, two);
 235        }
 236        return strbuf_detach(&res, NULL);
 237}
 238
 239static const char *external_diff(void)
 240{
 241        static const char *external_diff_cmd = NULL;
 242        static int done_preparing = 0;
 243
 244        if (done_preparing)
 245                return external_diff_cmd;
 246        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 247        if (!external_diff_cmd)
 248                external_diff_cmd = external_diff_cmd_cfg;
 249        done_preparing = 1;
 250        return external_diff_cmd;
 251}
 252
 253static struct diff_tempfile {
 254        const char *name; /* filename external diff should read from */
 255        char hex[41];
 256        char mode[10];
 257        char tmp_path[PATH_MAX];
 258} diff_temp[2];
 259
 260typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 261
 262struct emit_callback {
 263        int color_diff;
 264        unsigned ws_rule;
 265        int blank_at_eof_in_preimage;
 266        int blank_at_eof_in_postimage;
 267        int lno_in_preimage;
 268        int lno_in_postimage;
 269        sane_truncate_fn truncate;
 270        const char **label_path;
 271        struct diff_words_data *diff_words;
 272        struct diff_options *opt;
 273        int *found_changesp;
 274        struct strbuf *header;
 275};
 276
 277static int count_lines(const char *data, int size)
 278{
 279        int count, ch, completely_empty = 1, nl_just_seen = 0;
 280        count = 0;
 281        while (0 < size--) {
 282                ch = *data++;
 283                if (ch == '\n') {
 284                        count++;
 285                        nl_just_seen = 1;
 286                        completely_empty = 0;
 287                }
 288                else {
 289                        nl_just_seen = 0;
 290                        completely_empty = 0;
 291                }
 292        }
 293        if (completely_empty)
 294                return 0;
 295        if (!nl_just_seen)
 296                count++; /* no trailing newline */
 297        return count;
 298}
 299
 300static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 301{
 302        if (!DIFF_FILE_VALID(one)) {
 303                mf->ptr = (char *)""; /* does not matter */
 304                mf->size = 0;
 305                return 0;
 306        }
 307        else if (diff_populate_filespec(one, 0))
 308                return -1;
 309
 310        mf->ptr = one->data;
 311        mf->size = one->size;
 312        return 0;
 313}
 314
 315/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 316static unsigned long diff_filespec_size(struct diff_filespec *one)
 317{
 318        if (!DIFF_FILE_VALID(one))
 319                return 0;
 320        diff_populate_filespec(one, 1);
 321        return one->size;
 322}
 323
 324static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 325{
 326        char *ptr = mf->ptr;
 327        long size = mf->size;
 328        int cnt = 0;
 329
 330        if (!size)
 331                return cnt;
 332        ptr += size - 1; /* pointing at the very end */
 333        if (*ptr != '\n')
 334                ; /* incomplete line */
 335        else
 336                ptr--; /* skip the last LF */
 337        while (mf->ptr < ptr) {
 338                char *prev_eol;
 339                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 340                        if (*prev_eol == '\n')
 341                                break;
 342                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 343                        break;
 344                cnt++;
 345                ptr = prev_eol - 1;
 346        }
 347        return cnt;
 348}
 349
 350static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 351                               struct emit_callback *ecbdata)
 352{
 353        int l1, l2, at;
 354        unsigned ws_rule = ecbdata->ws_rule;
 355        l1 = count_trailing_blank(mf1, ws_rule);
 356        l2 = count_trailing_blank(mf2, ws_rule);
 357        if (l2 <= l1) {
 358                ecbdata->blank_at_eof_in_preimage = 0;
 359                ecbdata->blank_at_eof_in_postimage = 0;
 360                return;
 361        }
 362        at = count_lines(mf1->ptr, mf1->size);
 363        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 364
 365        at = count_lines(mf2->ptr, mf2->size);
 366        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 367}
 368
 369static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 370                        int first, const char *line, int len)
 371{
 372        int has_trailing_newline, has_trailing_carriage_return;
 373        int nofirst;
 374        FILE *file = o->file;
 375
 376        if (o->output_prefix) {
 377                struct strbuf *msg = NULL;
 378                msg = o->output_prefix(o, o->output_prefix_data);
 379                assert(msg);
 380                fwrite(msg->buf, msg->len, 1, file);
 381        }
 382
 383        if (len == 0) {
 384                has_trailing_newline = (first == '\n');
 385                has_trailing_carriage_return = (!has_trailing_newline &&
 386                                                (first == '\r'));
 387                nofirst = has_trailing_newline || has_trailing_carriage_return;
 388        } else {
 389                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 390                if (has_trailing_newline)
 391                        len--;
 392                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 393                if (has_trailing_carriage_return)
 394                        len--;
 395                nofirst = 0;
 396        }
 397
 398        if (len || !nofirst) {
 399                fputs(set, file);
 400                if (!nofirst)
 401                        fputc(first, file);
 402                fwrite(line, len, 1, file);
 403                fputs(reset, file);
 404        }
 405        if (has_trailing_carriage_return)
 406                fputc('\r', file);
 407        if (has_trailing_newline)
 408                fputc('\n', file);
 409}
 410
 411static void emit_line(struct diff_options *o, const char *set, const char *reset,
 412                      const char *line, int len)
 413{
 414        emit_line_0(o, set, reset, line[0], line+1, len-1);
 415}
 416
 417static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 418{
 419        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 420              ecbdata->blank_at_eof_in_preimage &&
 421              ecbdata->blank_at_eof_in_postimage &&
 422              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 423              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 424                return 0;
 425        return ws_blank_line(line, len, ecbdata->ws_rule);
 426}
 427
 428static void emit_add_line(const char *reset,
 429                          struct emit_callback *ecbdata,
 430                          const char *line, int len)
 431{
 432        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 433        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 434
 435        if (!*ws)
 436                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 437        else if (new_blank_line_at_eof(ecbdata, line, len))
 438                /* Blank line at EOF - paint '+' as well */
 439                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 440        else {
 441                /* Emit just the prefix, then the rest. */
 442                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 443                ws_check_emit(line, len, ecbdata->ws_rule,
 444                              ecbdata->opt->file, set, reset, ws);
 445        }
 446}
 447
 448static void emit_hunk_header(struct emit_callback *ecbdata,
 449                             const char *line, int len)
 450{
 451        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 452        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 453        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 454        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 455        static const char atat[2] = { '@', '@' };
 456        const char *cp, *ep;
 457        struct strbuf msgbuf = STRBUF_INIT;
 458        int org_len = len;
 459        int i = 1;
 460
 461        /*
 462         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 463         * it always is at least 10 bytes long.
 464         */
 465        if (len < 10 ||
 466            memcmp(line, atat, 2) ||
 467            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 468                emit_line(ecbdata->opt, plain, reset, line, len);
 469                return;
 470        }
 471        ep += 2; /* skip over @@ */
 472
 473        /* The hunk header in fraginfo color */
 474        strbuf_add(&msgbuf, frag, strlen(frag));
 475        strbuf_add(&msgbuf, line, ep - line);
 476        strbuf_add(&msgbuf, reset, strlen(reset));
 477
 478        /*
 479         * trailing "\r\n"
 480         */
 481        for ( ; i < 3; i++)
 482                if (line[len - i] == '\r' || line[len - i] == '\n')
 483                        len--;
 484
 485        /* blank before the func header */
 486        for (cp = ep; ep - line < len; ep++)
 487                if (*ep != ' ' && *ep != '\t')
 488                        break;
 489        if (ep != cp) {
 490                strbuf_add(&msgbuf, plain, strlen(plain));
 491                strbuf_add(&msgbuf, cp, ep - cp);
 492                strbuf_add(&msgbuf, reset, strlen(reset));
 493        }
 494
 495        if (ep < line + len) {
 496                strbuf_add(&msgbuf, func, strlen(func));
 497                strbuf_add(&msgbuf, ep, line + len - ep);
 498                strbuf_add(&msgbuf, reset, strlen(reset));
 499        }
 500
 501        strbuf_add(&msgbuf, line + len, org_len - len);
 502        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 503        strbuf_release(&msgbuf);
 504}
 505
 506static struct diff_tempfile *claim_diff_tempfile(void) {
 507        int i;
 508        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 509                if (!diff_temp[i].name)
 510                        return diff_temp + i;
 511        die("BUG: diff is failing to clean up its tempfiles");
 512}
 513
 514static int remove_tempfile_installed;
 515
 516static void remove_tempfile(void)
 517{
 518        int i;
 519        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 520                if (diff_temp[i].name == diff_temp[i].tmp_path)
 521                        unlink_or_warn(diff_temp[i].name);
 522                diff_temp[i].name = NULL;
 523        }
 524}
 525
 526static void remove_tempfile_on_signal(int signo)
 527{
 528        remove_tempfile();
 529        sigchain_pop(signo);
 530        raise(signo);
 531}
 532
 533static void print_line_count(FILE *file, int count)
 534{
 535        switch (count) {
 536        case 0:
 537                fprintf(file, "0,0");
 538                break;
 539        case 1:
 540                fprintf(file, "1");
 541                break;
 542        default:
 543                fprintf(file, "1,%d", count);
 544                break;
 545        }
 546}
 547
 548static void emit_rewrite_lines(struct emit_callback *ecb,
 549                               int prefix, const char *data, int size)
 550{
 551        const char *endp = NULL;
 552        static const char *nneof = " No newline at end of file\n";
 553        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 554        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 555
 556        while (0 < size) {
 557                int len;
 558
 559                endp = memchr(data, '\n', size);
 560                len = endp ? (endp - data + 1) : size;
 561                if (prefix != '+') {
 562                        ecb->lno_in_preimage++;
 563                        emit_line_0(ecb->opt, old, reset, '-',
 564                                    data, len);
 565                } else {
 566                        ecb->lno_in_postimage++;
 567                        emit_add_line(reset, ecb, data, len);
 568                }
 569                size -= len;
 570                data += len;
 571        }
 572        if (!endp) {
 573                const char *plain = diff_get_color(ecb->color_diff,
 574                                                   DIFF_PLAIN);
 575                emit_line_0(ecb->opt, plain, reset, '\\',
 576                            nneof, strlen(nneof));
 577        }
 578}
 579
 580static void emit_rewrite_diff(const char *name_a,
 581                              const char *name_b,
 582                              struct diff_filespec *one,
 583                              struct diff_filespec *two,
 584                              struct userdiff_driver *textconv_one,
 585                              struct userdiff_driver *textconv_two,
 586                              struct diff_options *o)
 587{
 588        int lc_a, lc_b;
 589        const char *name_a_tab, *name_b_tab;
 590        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 591        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 592        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 593        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 594        const char *a_prefix, *b_prefix;
 595        char *data_one, *data_two;
 596        size_t size_one, size_two;
 597        struct emit_callback ecbdata;
 598        char *line_prefix = "";
 599        struct strbuf *msgbuf;
 600
 601        if (o && o->output_prefix) {
 602                msgbuf = o->output_prefix(o, o->output_prefix_data);
 603                line_prefix = msgbuf->buf;
 604        }
 605
 606        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 607                a_prefix = o->b_prefix;
 608                b_prefix = o->a_prefix;
 609        } else {
 610                a_prefix = o->a_prefix;
 611                b_prefix = o->b_prefix;
 612        }
 613
 614        name_a += (*name_a == '/');
 615        name_b += (*name_b == '/');
 616        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 617        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 618
 619        strbuf_reset(&a_name);
 620        strbuf_reset(&b_name);
 621        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 622        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 623
 624        size_one = fill_textconv(textconv_one, one, &data_one);
 625        size_two = fill_textconv(textconv_two, two, &data_two);
 626
 627        memset(&ecbdata, 0, sizeof(ecbdata));
 628        ecbdata.color_diff = want_color(o->use_color);
 629        ecbdata.found_changesp = &o->found_changes;
 630        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 631        ecbdata.opt = o;
 632        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 633                mmfile_t mf1, mf2;
 634                mf1.ptr = (char *)data_one;
 635                mf2.ptr = (char *)data_two;
 636                mf1.size = size_one;
 637                mf2.size = size_two;
 638                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 639        }
 640        ecbdata.lno_in_preimage = 1;
 641        ecbdata.lno_in_postimage = 1;
 642
 643        lc_a = count_lines(data_one, size_one);
 644        lc_b = count_lines(data_two, size_two);
 645        fprintf(o->file,
 646                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 647                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 648                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 649                line_prefix, fraginfo);
 650        if (!o->irreversible_delete)
 651                print_line_count(o->file, lc_a);
 652        else
 653                fprintf(o->file, "?,?");
 654        fprintf(o->file, " +");
 655        print_line_count(o->file, lc_b);
 656        fprintf(o->file, " @@%s\n", reset);
 657        if (lc_a && !o->irreversible_delete)
 658                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 659        if (lc_b)
 660                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 661        if (textconv_one)
 662                free((char *)data_one);
 663        if (textconv_two)
 664                free((char *)data_two);
 665}
 666
 667struct diff_words_buffer {
 668        mmfile_t text;
 669        long alloc;
 670        struct diff_words_orig {
 671                const char *begin, *end;
 672        } *orig;
 673        int orig_nr, orig_alloc;
 674};
 675
 676static void diff_words_append(char *line, unsigned long len,
 677                struct diff_words_buffer *buffer)
 678{
 679        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 680        line++;
 681        len--;
 682        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 683        buffer->text.size += len;
 684        buffer->text.ptr[buffer->text.size] = '\0';
 685}
 686
 687struct diff_words_style_elem {
 688        const char *prefix;
 689        const char *suffix;
 690        const char *color; /* NULL; filled in by the setup code if
 691                            * color is enabled */
 692};
 693
 694struct diff_words_style {
 695        enum diff_words_type type;
 696        struct diff_words_style_elem new, old, ctx;
 697        const char *newline;
 698};
 699
 700static struct diff_words_style diff_words_styles[] = {
 701        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 702        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 703        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 704};
 705
 706struct diff_words_data {
 707        struct diff_words_buffer minus, plus;
 708        const char *current_plus;
 709        int last_minus;
 710        struct diff_options *opt;
 711        regex_t *word_regex;
 712        enum diff_words_type type;
 713        struct diff_words_style *style;
 714};
 715
 716static int fn_out_diff_words_write_helper(FILE *fp,
 717                                          struct diff_words_style_elem *st_el,
 718                                          const char *newline,
 719                                          size_t count, const char *buf,
 720                                          const char *line_prefix)
 721{
 722        int print = 0;
 723
 724        while (count) {
 725                char *p = memchr(buf, '\n', count);
 726                if (print)
 727                        fputs(line_prefix, fp);
 728                if (p != buf) {
 729                        if (st_el->color && fputs(st_el->color, fp) < 0)
 730                                return -1;
 731                        if (fputs(st_el->prefix, fp) < 0 ||
 732                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 733                            fputs(st_el->suffix, fp) < 0)
 734                                return -1;
 735                        if (st_el->color && *st_el->color
 736                            && fputs(GIT_COLOR_RESET, fp) < 0)
 737                                return -1;
 738                }
 739                if (!p)
 740                        return 0;
 741                if (fputs(newline, fp) < 0)
 742                        return -1;
 743                count -= p + 1 - buf;
 744                buf = p + 1;
 745                print = 1;
 746        }
 747        return 0;
 748}
 749
 750/*
 751 * '--color-words' algorithm can be described as:
 752 *
 753 *   1. collect a the minus/plus lines of a diff hunk, divided into
 754 *      minus-lines and plus-lines;
 755 *
 756 *   2. break both minus-lines and plus-lines into words and
 757 *      place them into two mmfile_t with one word for each line;
 758 *
 759 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 760 *
 761 * And for the common parts of the both file, we output the plus side text.
 762 * diff_words->current_plus is used to trace the current position of the plus file
 763 * which printed. diff_words->last_minus is used to trace the last minus word
 764 * printed.
 765 *
 766 * For '--graph' to work with '--color-words', we need to output the graph prefix
 767 * on each line of color words output. Generally, there are two conditions on
 768 * which we should output the prefix.
 769 *
 770 *   1. diff_words->last_minus == 0 &&
 771 *      diff_words->current_plus == diff_words->plus.text.ptr
 772 *
 773 *      that is: the plus text must start as a new line, and if there is no minus
 774 *      word printed, a graph prefix must be printed.
 775 *
 776 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 777 *      *(diff_words->current_plus - 1) == '\n'
 778 *
 779 *      that is: a graph prefix must be printed following a '\n'
 780 */
 781static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 782{
 783        if ((diff_words->last_minus == 0 &&
 784                diff_words->current_plus == diff_words->plus.text.ptr) ||
 785                (diff_words->current_plus > diff_words->plus.text.ptr &&
 786                *(diff_words->current_plus - 1) == '\n')) {
 787                return 1;
 788        } else {
 789                return 0;
 790        }
 791}
 792
 793static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 794{
 795        struct diff_words_data *diff_words = priv;
 796        struct diff_words_style *style = diff_words->style;
 797        int minus_first, minus_len, plus_first, plus_len;
 798        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 799        struct diff_options *opt = diff_words->opt;
 800        struct strbuf *msgbuf;
 801        char *line_prefix = "";
 802
 803        if (line[0] != '@' || parse_hunk_header(line, len,
 804                        &minus_first, &minus_len, &plus_first, &plus_len))
 805                return;
 806
 807        assert(opt);
 808        if (opt->output_prefix) {
 809                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 810                line_prefix = msgbuf->buf;
 811        }
 812
 813        /* POSIX requires that first be decremented by one if len == 0... */
 814        if (minus_len) {
 815                minus_begin = diff_words->minus.orig[minus_first].begin;
 816                minus_end =
 817                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 818        } else
 819                minus_begin = minus_end =
 820                        diff_words->minus.orig[minus_first].end;
 821
 822        if (plus_len) {
 823                plus_begin = diff_words->plus.orig[plus_first].begin;
 824                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 825        } else
 826                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 827
 828        if (color_words_output_graph_prefix(diff_words)) {
 829                fputs(line_prefix, diff_words->opt->file);
 830        }
 831        if (diff_words->current_plus != plus_begin) {
 832                fn_out_diff_words_write_helper(diff_words->opt->file,
 833                                &style->ctx, style->newline,
 834                                plus_begin - diff_words->current_plus,
 835                                diff_words->current_plus, line_prefix);
 836                if (*(plus_begin - 1) == '\n')
 837                        fputs(line_prefix, diff_words->opt->file);
 838        }
 839        if (minus_begin != minus_end) {
 840                fn_out_diff_words_write_helper(diff_words->opt->file,
 841                                &style->old, style->newline,
 842                                minus_end - minus_begin, minus_begin,
 843                                line_prefix);
 844        }
 845        if (plus_begin != plus_end) {
 846                fn_out_diff_words_write_helper(diff_words->opt->file,
 847                                &style->new, style->newline,
 848                                plus_end - plus_begin, plus_begin,
 849                                line_prefix);
 850        }
 851
 852        diff_words->current_plus = plus_end;
 853        diff_words->last_minus = minus_first;
 854}
 855
 856/* This function starts looking at *begin, and returns 0 iff a word was found. */
 857static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 858                int *begin, int *end)
 859{
 860        if (word_regex && *begin < buffer->size) {
 861                regmatch_t match[1];
 862                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 863                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 864                                        '\n', match[0].rm_eo - match[0].rm_so);
 865                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 866                        *begin += match[0].rm_so;
 867                        return *begin >= *end;
 868                }
 869                return -1;
 870        }
 871
 872        /* find the next word */
 873        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 874                (*begin)++;
 875        if (*begin >= buffer->size)
 876                return -1;
 877
 878        /* find the end of the word */
 879        *end = *begin + 1;
 880        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 881                (*end)++;
 882
 883        return 0;
 884}
 885
 886/*
 887 * This function splits the words in buffer->text, stores the list with
 888 * newline separator into out, and saves the offsets of the original words
 889 * in buffer->orig.
 890 */
 891static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 892                regex_t *word_regex)
 893{
 894        int i, j;
 895        long alloc = 0;
 896
 897        out->size = 0;
 898        out->ptr = NULL;
 899
 900        /* fake an empty "0th" word */
 901        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 902        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 903        buffer->orig_nr = 1;
 904
 905        for (i = 0; i < buffer->text.size; i++) {
 906                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 907                        return;
 908
 909                /* store original boundaries */
 910                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 911                                buffer->orig_alloc);
 912                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 913                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 914                buffer->orig_nr++;
 915
 916                /* store one word */
 917                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 918                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 919                out->ptr[out->size + j - i] = '\n';
 920                out->size += j - i + 1;
 921
 922                i = j - 1;
 923        }
 924}
 925
 926/* this executes the word diff on the accumulated buffers */
 927static void diff_words_show(struct diff_words_data *diff_words)
 928{
 929        xpparam_t xpp;
 930        xdemitconf_t xecfg;
 931        mmfile_t minus, plus;
 932        struct diff_words_style *style = diff_words->style;
 933
 934        struct diff_options *opt = diff_words->opt;
 935        struct strbuf *msgbuf;
 936        char *line_prefix = "";
 937
 938        assert(opt);
 939        if (opt->output_prefix) {
 940                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 941                line_prefix = msgbuf->buf;
 942        }
 943
 944        /* special case: only removal */
 945        if (!diff_words->plus.text.size) {
 946                fputs(line_prefix, diff_words->opt->file);
 947                fn_out_diff_words_write_helper(diff_words->opt->file,
 948                        &style->old, style->newline,
 949                        diff_words->minus.text.size,
 950                        diff_words->minus.text.ptr, line_prefix);
 951                diff_words->minus.text.size = 0;
 952                return;
 953        }
 954
 955        diff_words->current_plus = diff_words->plus.text.ptr;
 956        diff_words->last_minus = 0;
 957
 958        memset(&xpp, 0, sizeof(xpp));
 959        memset(&xecfg, 0, sizeof(xecfg));
 960        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 961        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 962        xpp.flags = 0;
 963        /* as only the hunk header will be parsed, we need a 0-context */
 964        xecfg.ctxlen = 0;
 965        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 966                      &xpp, &xecfg);
 967        free(minus.ptr);
 968        free(plus.ptr);
 969        if (diff_words->current_plus != diff_words->plus.text.ptr +
 970                        diff_words->plus.text.size) {
 971                if (color_words_output_graph_prefix(diff_words))
 972                        fputs(line_prefix, diff_words->opt->file);
 973                fn_out_diff_words_write_helper(diff_words->opt->file,
 974                        &style->ctx, style->newline,
 975                        diff_words->plus.text.ptr + diff_words->plus.text.size
 976                        - diff_words->current_plus, diff_words->current_plus,
 977                        line_prefix);
 978        }
 979        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 980}
 981
 982/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 983static void diff_words_flush(struct emit_callback *ecbdata)
 984{
 985        if (ecbdata->diff_words->minus.text.size ||
 986            ecbdata->diff_words->plus.text.size)
 987                diff_words_show(ecbdata->diff_words);
 988}
 989
 990static void diff_filespec_load_driver(struct diff_filespec *one)
 991{
 992        /* Use already-loaded driver */
 993        if (one->driver)
 994                return;
 995
 996        if (S_ISREG(one->mode))
 997                one->driver = userdiff_find_by_path(one->path);
 998
 999        /* Fallback to default settings */
1000        if (!one->driver)
1001                one->driver = userdiff_find_by_name("default");
1002}
1003
1004static const char *userdiff_word_regex(struct diff_filespec *one)
1005{
1006        diff_filespec_load_driver(one);
1007        return one->driver->word_regex;
1008}
1009
1010static void init_diff_words_data(struct emit_callback *ecbdata,
1011                                 struct diff_options *orig_opts,
1012                                 struct diff_filespec *one,
1013                                 struct diff_filespec *two)
1014{
1015        int i;
1016        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1017        memcpy(o, orig_opts, sizeof(struct diff_options));
1018
1019        ecbdata->diff_words =
1020                xcalloc(1, sizeof(struct diff_words_data));
1021        ecbdata->diff_words->type = o->word_diff;
1022        ecbdata->diff_words->opt = o;
1023        if (!o->word_regex)
1024                o->word_regex = userdiff_word_regex(one);
1025        if (!o->word_regex)
1026                o->word_regex = userdiff_word_regex(two);
1027        if (!o->word_regex)
1028                o->word_regex = diff_word_regex_cfg;
1029        if (o->word_regex) {
1030                ecbdata->diff_words->word_regex = (regex_t *)
1031                        xmalloc(sizeof(regex_t));
1032                if (regcomp(ecbdata->diff_words->word_regex,
1033                            o->word_regex,
1034                            REG_EXTENDED | REG_NEWLINE))
1035                        die ("Invalid regular expression: %s",
1036                             o->word_regex);
1037        }
1038        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1039                if (o->word_diff == diff_words_styles[i].type) {
1040                        ecbdata->diff_words->style =
1041                                &diff_words_styles[i];
1042                        break;
1043                }
1044        }
1045        if (want_color(o->use_color)) {
1046                struct diff_words_style *st = ecbdata->diff_words->style;
1047                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1048                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1049                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1050        }
1051}
1052
1053static void free_diff_words_data(struct emit_callback *ecbdata)
1054{
1055        if (ecbdata->diff_words) {
1056                diff_words_flush(ecbdata);
1057                free (ecbdata->diff_words->opt);
1058                free (ecbdata->diff_words->minus.text.ptr);
1059                free (ecbdata->diff_words->minus.orig);
1060                free (ecbdata->diff_words->plus.text.ptr);
1061                free (ecbdata->diff_words->plus.orig);
1062                if (ecbdata->diff_words->word_regex) {
1063                        regfree(ecbdata->diff_words->word_regex);
1064                        free(ecbdata->diff_words->word_regex);
1065                }
1066                free(ecbdata->diff_words);
1067                ecbdata->diff_words = NULL;
1068        }
1069}
1070
1071const char *diff_get_color(int diff_use_color, enum color_diff ix)
1072{
1073        if (want_color(diff_use_color))
1074                return diff_colors[ix];
1075        return "";
1076}
1077
1078static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1079{
1080        const char *cp;
1081        unsigned long allot;
1082        size_t l = len;
1083
1084        if (ecb->truncate)
1085                return ecb->truncate(line, len);
1086        cp = line;
1087        allot = l;
1088        while (0 < l) {
1089                (void) utf8_width(&cp, &l);
1090                if (!cp)
1091                        break; /* truncated in the middle? */
1092        }
1093        return allot - l;
1094}
1095
1096static void find_lno(const char *line, struct emit_callback *ecbdata)
1097{
1098        const char *p;
1099        ecbdata->lno_in_preimage = 0;
1100        ecbdata->lno_in_postimage = 0;
1101        p = strchr(line, '-');
1102        if (!p)
1103                return; /* cannot happen */
1104        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1105        p = strchr(p, '+');
1106        if (!p)
1107                return; /* cannot happen */
1108        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1109}
1110
1111static void fn_out_consume(void *priv, char *line, unsigned long len)
1112{
1113        struct emit_callback *ecbdata = priv;
1114        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1115        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1116        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1117        struct diff_options *o = ecbdata->opt;
1118        char *line_prefix = "";
1119        struct strbuf *msgbuf;
1120
1121        if (o && o->output_prefix) {
1122                msgbuf = o->output_prefix(o, o->output_prefix_data);
1123                line_prefix = msgbuf->buf;
1124        }
1125
1126        if (ecbdata->header) {
1127                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1128                strbuf_reset(ecbdata->header);
1129                ecbdata->header = NULL;
1130        }
1131        *(ecbdata->found_changesp) = 1;
1132
1133        if (ecbdata->label_path[0]) {
1134                const char *name_a_tab, *name_b_tab;
1135
1136                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1137                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1138
1139                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1140                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1141                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1142                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1143                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1144        }
1145
1146        if (diff_suppress_blank_empty
1147            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1148                line[0] = '\n';
1149                len = 1;
1150        }
1151
1152        if (line[0] == '@') {
1153                if (ecbdata->diff_words)
1154                        diff_words_flush(ecbdata);
1155                len = sane_truncate_line(ecbdata, line, len);
1156                find_lno(line, ecbdata);
1157                emit_hunk_header(ecbdata, line, len);
1158                if (line[len-1] != '\n')
1159                        putc('\n', ecbdata->opt->file);
1160                return;
1161        }
1162
1163        if (len < 1) {
1164                emit_line(ecbdata->opt, reset, reset, line, len);
1165                if (ecbdata->diff_words
1166                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1167                        fputs("~\n", ecbdata->opt->file);
1168                return;
1169        }
1170
1171        if (ecbdata->diff_words) {
1172                if (line[0] == '-') {
1173                        diff_words_append(line, len,
1174                                          &ecbdata->diff_words->minus);
1175                        return;
1176                } else if (line[0] == '+') {
1177                        diff_words_append(line, len,
1178                                          &ecbdata->diff_words->plus);
1179                        return;
1180                } else if (!prefixcmp(line, "\\ ")) {
1181                        /*
1182                         * Eat the "no newline at eof" marker as if we
1183                         * saw a "+" or "-" line with nothing on it,
1184                         * and return without diff_words_flush() to
1185                         * defer processing. If this is the end of
1186                         * preimage, more "+" lines may come after it.
1187                         */
1188                        return;
1189                }
1190                diff_words_flush(ecbdata);
1191                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1192                        emit_line(ecbdata->opt, plain, reset, line, len);
1193                        fputs("~\n", ecbdata->opt->file);
1194                } else {
1195                        /*
1196                         * Skip the prefix character, if any.  With
1197                         * diff_suppress_blank_empty, there may be
1198                         * none.
1199                         */
1200                        if (line[0] != '\n') {
1201                              line++;
1202                              len--;
1203                        }
1204                        emit_line(ecbdata->opt, plain, reset, line, len);
1205                }
1206                return;
1207        }
1208
1209        if (line[0] != '+') {
1210                const char *color =
1211                        diff_get_color(ecbdata->color_diff,
1212                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1213                ecbdata->lno_in_preimage++;
1214                if (line[0] == ' ')
1215                        ecbdata->lno_in_postimage++;
1216                emit_line(ecbdata->opt, color, reset, line, len);
1217        } else {
1218                ecbdata->lno_in_postimage++;
1219                emit_add_line(reset, ecbdata, line + 1, len - 1);
1220        }
1221}
1222
1223static char *pprint_rename(const char *a, const char *b)
1224{
1225        const char *old = a;
1226        const char *new = b;
1227        struct strbuf name = STRBUF_INIT;
1228        int pfx_length, sfx_length;
1229        int len_a = strlen(a);
1230        int len_b = strlen(b);
1231        int a_midlen, b_midlen;
1232        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1233        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1234
1235        if (qlen_a || qlen_b) {
1236                quote_c_style(a, &name, NULL, 0);
1237                strbuf_addstr(&name, " => ");
1238                quote_c_style(b, &name, NULL, 0);
1239                return strbuf_detach(&name, NULL);
1240        }
1241
1242        /* Find common prefix */
1243        pfx_length = 0;
1244        while (*old && *new && *old == *new) {
1245                if (*old == '/')
1246                        pfx_length = old - a + 1;
1247                old++;
1248                new++;
1249        }
1250
1251        /* Find common suffix */
1252        old = a + len_a;
1253        new = b + len_b;
1254        sfx_length = 0;
1255        while (a <= old && b <= new && *old == *new) {
1256                if (*old == '/')
1257                        sfx_length = len_a - (old - a);
1258                old--;
1259                new--;
1260        }
1261
1262        /*
1263         * pfx{mid-a => mid-b}sfx
1264         * {pfx-a => pfx-b}sfx
1265         * pfx{sfx-a => sfx-b}
1266         * name-a => name-b
1267         */
1268        a_midlen = len_a - pfx_length - sfx_length;
1269        b_midlen = len_b - pfx_length - sfx_length;
1270        if (a_midlen < 0)
1271                a_midlen = 0;
1272        if (b_midlen < 0)
1273                b_midlen = 0;
1274
1275        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1276        if (pfx_length + sfx_length) {
1277                strbuf_add(&name, a, pfx_length);
1278                strbuf_addch(&name, '{');
1279        }
1280        strbuf_add(&name, a + pfx_length, a_midlen);
1281        strbuf_addstr(&name, " => ");
1282        strbuf_add(&name, b + pfx_length, b_midlen);
1283        if (pfx_length + sfx_length) {
1284                strbuf_addch(&name, '}');
1285                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1286        }
1287        return strbuf_detach(&name, NULL);
1288}
1289
1290struct diffstat_t {
1291        int nr;
1292        int alloc;
1293        struct diffstat_file {
1294                char *from_name;
1295                char *name;
1296                char *print_name;
1297                unsigned is_unmerged:1;
1298                unsigned is_binary:1;
1299                unsigned is_renamed:1;
1300                uintmax_t added, deleted;
1301        } **files;
1302};
1303
1304static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1305                                          const char *name_a,
1306                                          const char *name_b)
1307{
1308        struct diffstat_file *x;
1309        x = xcalloc(sizeof (*x), 1);
1310        if (diffstat->nr == diffstat->alloc) {
1311                diffstat->alloc = alloc_nr(diffstat->alloc);
1312                diffstat->files = xrealloc(diffstat->files,
1313                                diffstat->alloc * sizeof(x));
1314        }
1315        diffstat->files[diffstat->nr++] = x;
1316        if (name_b) {
1317                x->from_name = xstrdup(name_a);
1318                x->name = xstrdup(name_b);
1319                x->is_renamed = 1;
1320        }
1321        else {
1322                x->from_name = NULL;
1323                x->name = xstrdup(name_a);
1324        }
1325        return x;
1326}
1327
1328static void diffstat_consume(void *priv, char *line, unsigned long len)
1329{
1330        struct diffstat_t *diffstat = priv;
1331        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1332
1333        if (line[0] == '+')
1334                x->added++;
1335        else if (line[0] == '-')
1336                x->deleted++;
1337}
1338
1339const char mime_boundary_leader[] = "------------";
1340
1341static int scale_linear(int it, int width, int max_change)
1342{
1343        if (!it)
1344                return 0;
1345        /*
1346         * make sure that at least one '-' or '+' is printed if
1347         * there is any change to this path. The easiest way is to
1348         * scale linearly as if the alloted width is one column shorter
1349         * than it is, and then add 1 to the result.
1350         */
1351        return 1 + (it * (width - 1) / max_change);
1352}
1353
1354static void show_name(FILE *file,
1355                      const char *prefix, const char *name, int len)
1356{
1357        fprintf(file, " %s%-*s |", prefix, len, name);
1358}
1359
1360static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1361{
1362        if (cnt <= 0)
1363                return;
1364        fprintf(file, "%s", set);
1365        while (cnt--)
1366                putc(ch, file);
1367        fprintf(file, "%s", reset);
1368}
1369
1370static void fill_print_name(struct diffstat_file *file)
1371{
1372        char *pname;
1373
1374        if (file->print_name)
1375                return;
1376
1377        if (!file->is_renamed) {
1378                struct strbuf buf = STRBUF_INIT;
1379                if (quote_c_style(file->name, &buf, NULL, 0)) {
1380                        pname = strbuf_detach(&buf, NULL);
1381                } else {
1382                        pname = file->name;
1383                        strbuf_release(&buf);
1384                }
1385        } else {
1386                pname = pprint_rename(file->from_name, file->name);
1387        }
1388        file->print_name = pname;
1389}
1390
1391int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1392{
1393        struct strbuf sb = STRBUF_INIT;
1394        int ret;
1395
1396        if (!files) {
1397                assert(insertions == 0 && deletions == 0);
1398                return fputs(_(" 0 files changed\n"), fp);
1399        }
1400
1401        strbuf_addf(&sb,
1402                    Q_(" %d file changed", " %d files changed", files),
1403                    files);
1404
1405        /*
1406         * For binary diff, the caller may want to print "x files
1407         * changed" with insertions == 0 && deletions == 0.
1408         *
1409         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1410         * is probably less confusing (i.e skip over "2 files changed
1411         * but nothing about added/removed lines? Is this a bug in Git?").
1412         */
1413        if (insertions || deletions == 0) {
1414                /*
1415                 * TRANSLATORS: "+" in (+) is a line addition marker;
1416                 * do not translate it.
1417                 */
1418                strbuf_addf(&sb,
1419                            Q_(", %d insertion(+)", ", %d insertions(+)",
1420                               insertions),
1421                            insertions);
1422        }
1423
1424        if (deletions || insertions == 0) {
1425                /*
1426                 * TRANSLATORS: "-" in (-) is a line removal marker;
1427                 * do not translate it.
1428                 */
1429                strbuf_addf(&sb,
1430                            Q_(", %d deletion(-)", ", %d deletions(-)",
1431                               deletions),
1432                            deletions);
1433        }
1434        strbuf_addch(&sb, '\n');
1435        ret = fputs(sb.buf, fp);
1436        strbuf_release(&sb);
1437        return ret;
1438}
1439
1440static void show_stats(struct diffstat_t *data, struct diff_options *options)
1441{
1442        int i, len, add, del, adds = 0, dels = 0;
1443        uintmax_t max_change = 0, max_len = 0;
1444        int total_files = data->nr;
1445        int width, name_width, count;
1446        const char *reset, *add_c, *del_c;
1447        const char *line_prefix = "";
1448        int extra_shown = 0;
1449        struct strbuf *msg = NULL;
1450
1451        if (data->nr == 0)
1452                return;
1453
1454        if (options->output_prefix) {
1455                msg = options->output_prefix(options, options->output_prefix_data);
1456                line_prefix = msg->buf;
1457        }
1458
1459        width = options->stat_width ? options->stat_width : 80;
1460        name_width = options->stat_name_width ? options->stat_name_width : 50;
1461        count = options->stat_count ? options->stat_count : data->nr;
1462
1463        /* Sanity: give at least 5 columns to the graph,
1464         * but leave at least 10 columns for the name.
1465         */
1466        if (width < 25)
1467                width = 25;
1468        if (name_width < 10)
1469                name_width = 10;
1470        else if (width < name_width + 15)
1471                name_width = width - 15;
1472
1473        /* Find the longest filename and max number of changes */
1474        reset = diff_get_color_opt(options, DIFF_RESET);
1475        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1476        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1477
1478        for (i = 0; (i < count) && (i < data->nr); i++) {
1479                struct diffstat_file *file = data->files[i];
1480                uintmax_t change = file->added + file->deleted;
1481                if (!data->files[i]->is_renamed &&
1482                         (change == 0)) {
1483                        count++; /* not shown == room for one more */
1484                        continue;
1485                }
1486                fill_print_name(file);
1487                len = strlen(file->print_name);
1488                if (max_len < len)
1489                        max_len = len;
1490
1491                if (file->is_binary || file->is_unmerged)
1492                        continue;
1493                if (max_change < change)
1494                        max_change = change;
1495        }
1496        count = i; /* min(count, data->nr) */
1497
1498        /* Compute the width of the graph part;
1499         * 10 is for one blank at the beginning of the line plus
1500         * " | count " between the name and the graph.
1501         *
1502         * From here on, name_width is the width of the name area,
1503         * and width is the width of the graph area.
1504         */
1505        name_width = (name_width < max_len) ? name_width : max_len;
1506        if (width < (name_width + 10) + max_change)
1507                width = width - (name_width + 10);
1508        else
1509                width = max_change;
1510
1511        for (i = 0; i < count; i++) {
1512                const char *prefix = "";
1513                char *name = data->files[i]->print_name;
1514                uintmax_t added = data->files[i]->added;
1515                uintmax_t deleted = data->files[i]->deleted;
1516                int name_len;
1517
1518                if (!data->files[i]->is_renamed &&
1519                         (added + deleted == 0)) {
1520                        total_files--;
1521                        continue;
1522                }
1523                /*
1524                 * "scale" the filename
1525                 */
1526                len = name_width;
1527                name_len = strlen(name);
1528                if (name_width < name_len) {
1529                        char *slash;
1530                        prefix = "...";
1531                        len -= 3;
1532                        name += name_len - len;
1533                        slash = strchr(name, '/');
1534                        if (slash)
1535                                name = slash;
1536                }
1537
1538                if (data->files[i]->is_binary) {
1539                        fprintf(options->file, "%s", line_prefix);
1540                        show_name(options->file, prefix, name, len);
1541                        fprintf(options->file, "  Bin ");
1542                        fprintf(options->file, "%s%"PRIuMAX"%s",
1543                                del_c, deleted, reset);
1544                        fprintf(options->file, " -> ");
1545                        fprintf(options->file, "%s%"PRIuMAX"%s",
1546                                add_c, added, reset);
1547                        fprintf(options->file, " bytes");
1548                        fprintf(options->file, "\n");
1549                        continue;
1550                }
1551                else if (data->files[i]->is_unmerged) {
1552                        fprintf(options->file, "%s", line_prefix);
1553                        show_name(options->file, prefix, name, len);
1554                        fprintf(options->file, "  Unmerged\n");
1555                        continue;
1556                }
1557
1558                /*
1559                 * scale the add/delete
1560                 */
1561                add = added;
1562                del = deleted;
1563                adds += add;
1564                dels += del;
1565
1566                if (width <= max_change) {
1567                        int total = add + del;
1568
1569                        total = scale_linear(add + del, width, max_change);
1570                        if (total < 2 && add && del)
1571                                /* width >= 2 due to the sanity check */
1572                                total = 2;
1573                        if (add < del) {
1574                                add = scale_linear(add, width, max_change);
1575                                del = total - add;
1576                        } else {
1577                                del = scale_linear(del, width, max_change);
1578                                add = total - del;
1579                        }
1580                }
1581                fprintf(options->file, "%s", line_prefix);
1582                show_name(options->file, prefix, name, len);
1583                fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1584                                added + deleted ? " " : "");
1585                show_graph(options->file, '+', add, add_c, reset);
1586                show_graph(options->file, '-', del, del_c, reset);
1587                fprintf(options->file, "\n");
1588        }
1589        for (i = count; i < data->nr; i++) {
1590                uintmax_t added = data->files[i]->added;
1591                uintmax_t deleted = data->files[i]->deleted;
1592                if (!data->files[i]->is_renamed &&
1593                         (added + deleted == 0)) {
1594                        total_files--;
1595                        continue;
1596                }
1597                adds += added;
1598                dels += deleted;
1599                if (!extra_shown)
1600                        fprintf(options->file, "%s ...\n", line_prefix);
1601                extra_shown = 1;
1602        }
1603        fprintf(options->file, "%s", line_prefix);
1604        print_stat_summary(options->file, total_files, adds, dels);
1605}
1606
1607static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1608{
1609        int i, adds = 0, dels = 0, total_files = data->nr;
1610
1611        if (data->nr == 0)
1612                return;
1613
1614        for (i = 0; i < data->nr; i++) {
1615                if (!data->files[i]->is_binary &&
1616                    !data->files[i]->is_unmerged) {
1617                        int added = data->files[i]->added;
1618                        int deleted= data->files[i]->deleted;
1619                        if (!data->files[i]->is_renamed &&
1620                            (added + deleted == 0)) {
1621                                total_files--;
1622                        } else {
1623                                adds += added;
1624                                dels += deleted;
1625                        }
1626                }
1627        }
1628        if (options->output_prefix) {
1629                struct strbuf *msg = NULL;
1630                msg = options->output_prefix(options,
1631                                options->output_prefix_data);
1632                fprintf(options->file, "%s", msg->buf);
1633        }
1634        print_stat_summary(options->file, total_files, adds, dels);
1635}
1636
1637static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1638{
1639        int i;
1640
1641        if (data->nr == 0)
1642                return;
1643
1644        for (i = 0; i < data->nr; i++) {
1645                struct diffstat_file *file = data->files[i];
1646
1647                if (options->output_prefix) {
1648                        struct strbuf *msg = NULL;
1649                        msg = options->output_prefix(options,
1650                                        options->output_prefix_data);
1651                        fprintf(options->file, "%s", msg->buf);
1652                }
1653
1654                if (file->is_binary)
1655                        fprintf(options->file, "-\t-\t");
1656                else
1657                        fprintf(options->file,
1658                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1659                                file->added, file->deleted);
1660                if (options->line_termination) {
1661                        fill_print_name(file);
1662                        if (!file->is_renamed)
1663                                write_name_quoted(file->name, options->file,
1664                                                  options->line_termination);
1665                        else {
1666                                fputs(file->print_name, options->file);
1667                                putc(options->line_termination, options->file);
1668                        }
1669                } else {
1670                        if (file->is_renamed) {
1671                                putc('\0', options->file);
1672                                write_name_quoted(file->from_name, options->file, '\0');
1673                        }
1674                        write_name_quoted(file->name, options->file, '\0');
1675                }
1676        }
1677}
1678
1679struct dirstat_file {
1680        const char *name;
1681        unsigned long changed;
1682};
1683
1684struct dirstat_dir {
1685        struct dirstat_file *files;
1686        int alloc, nr, permille, cumulative;
1687};
1688
1689static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1690                unsigned long changed, const char *base, int baselen)
1691{
1692        unsigned long this_dir = 0;
1693        unsigned int sources = 0;
1694        const char *line_prefix = "";
1695        struct strbuf *msg = NULL;
1696
1697        if (opt->output_prefix) {
1698                msg = opt->output_prefix(opt, opt->output_prefix_data);
1699                line_prefix = msg->buf;
1700        }
1701
1702        while (dir->nr) {
1703                struct dirstat_file *f = dir->files;
1704                int namelen = strlen(f->name);
1705                unsigned long this;
1706                char *slash;
1707
1708                if (namelen < baselen)
1709                        break;
1710                if (memcmp(f->name, base, baselen))
1711                        break;
1712                slash = strchr(f->name + baselen, '/');
1713                if (slash) {
1714                        int newbaselen = slash + 1 - f->name;
1715                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1716                        sources++;
1717                } else {
1718                        this = f->changed;
1719                        dir->files++;
1720                        dir->nr--;
1721                        sources += 2;
1722                }
1723                this_dir += this;
1724        }
1725
1726        /*
1727         * We don't report dirstat's for
1728         *  - the top level
1729         *  - or cases where everything came from a single directory
1730         *    under this directory (sources == 1).
1731         */
1732        if (baselen && sources != 1) {
1733                if (this_dir) {
1734                        int permille = this_dir * 1000 / changed;
1735                        if (permille >= dir->permille) {
1736                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1737                                        permille / 10, permille % 10, baselen, base);
1738                                if (!dir->cumulative)
1739                                        return 0;
1740                        }
1741                }
1742        }
1743        return this_dir;
1744}
1745
1746static int dirstat_compare(const void *_a, const void *_b)
1747{
1748        const struct dirstat_file *a = _a;
1749        const struct dirstat_file *b = _b;
1750        return strcmp(a->name, b->name);
1751}
1752
1753static void show_dirstat(struct diff_options *options)
1754{
1755        int i;
1756        unsigned long changed;
1757        struct dirstat_dir dir;
1758        struct diff_queue_struct *q = &diff_queued_diff;
1759
1760        dir.files = NULL;
1761        dir.alloc = 0;
1762        dir.nr = 0;
1763        dir.permille = options->dirstat_permille;
1764        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1765
1766        changed = 0;
1767        for (i = 0; i < q->nr; i++) {
1768                struct diff_filepair *p = q->queue[i];
1769                const char *name;
1770                unsigned long copied, added, damage;
1771                int content_changed;
1772
1773                name = p->two->path ? p->two->path : p->one->path;
1774
1775                if (p->one->sha1_valid && p->two->sha1_valid)
1776                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1777                else
1778                        content_changed = 1;
1779
1780                if (!content_changed) {
1781                        /*
1782                         * The SHA1 has not changed, so pre-/post-content is
1783                         * identical. We can therefore skip looking at the
1784                         * file contents altogether.
1785                         */
1786                        damage = 0;
1787                        goto found_damage;
1788                }
1789
1790                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1791                        /*
1792                         * In --dirstat-by-file mode, we don't really need to
1793                         * look at the actual file contents at all.
1794                         * The fact that the SHA1 changed is enough for us to
1795                         * add this file to the list of results
1796                         * (with each file contributing equal damage).
1797                         */
1798                        damage = 1;
1799                        goto found_damage;
1800                }
1801
1802                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1803                        diff_populate_filespec(p->one, 0);
1804                        diff_populate_filespec(p->two, 0);
1805                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1806                                               &copied, &added);
1807                        diff_free_filespec_data(p->one);
1808                        diff_free_filespec_data(p->two);
1809                } else if (DIFF_FILE_VALID(p->one)) {
1810                        diff_populate_filespec(p->one, 1);
1811                        copied = added = 0;
1812                        diff_free_filespec_data(p->one);
1813                } else if (DIFF_FILE_VALID(p->two)) {
1814                        diff_populate_filespec(p->two, 1);
1815                        copied = 0;
1816                        added = p->two->size;
1817                        diff_free_filespec_data(p->two);
1818                } else
1819                        continue;
1820
1821                /*
1822                 * Original minus copied is the removed material,
1823                 * added is the new material.  They are both damages
1824                 * made to the preimage.
1825                 * If the resulting damage is zero, we know that
1826                 * diffcore_count_changes() considers the two entries to
1827                 * be identical, but since content_changed is true, we
1828                 * know that there must have been _some_ kind of change,
1829                 * so we force all entries to have damage > 0.
1830                 */
1831                damage = (p->one->size - copied) + added;
1832                if (!damage)
1833                        damage = 1;
1834
1835found_damage:
1836                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1837                dir.files[dir.nr].name = name;
1838                dir.files[dir.nr].changed = damage;
1839                changed += damage;
1840                dir.nr++;
1841        }
1842
1843        /* This can happen even with many files, if everything was renames */
1844        if (!changed)
1845                return;
1846
1847        /* Show all directories with more than x% of the changes */
1848        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1849        gather_dirstat(options, &dir, changed, "", 0);
1850}
1851
1852static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1853{
1854        int i;
1855        unsigned long changed;
1856        struct dirstat_dir dir;
1857
1858        if (data->nr == 0)
1859                return;
1860
1861        dir.files = NULL;
1862        dir.alloc = 0;
1863        dir.nr = 0;
1864        dir.permille = options->dirstat_permille;
1865        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1866
1867        changed = 0;
1868        for (i = 0; i < data->nr; i++) {
1869                struct diffstat_file *file = data->files[i];
1870                unsigned long damage = file->added + file->deleted;
1871                if (file->is_binary)
1872                        /*
1873                         * binary files counts bytes, not lines. Must find some
1874                         * way to normalize binary bytes vs. textual lines.
1875                         * The following heuristic assumes that there are 64
1876                         * bytes per "line".
1877                         * This is stupid and ugly, but very cheap...
1878                         */
1879                        damage = (damage + 63) / 64;
1880                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1881                dir.files[dir.nr].name = file->name;
1882                dir.files[dir.nr].changed = damage;
1883                changed += damage;
1884                dir.nr++;
1885        }
1886
1887        /* This can happen even with many files, if everything was renames */
1888        if (!changed)
1889                return;
1890
1891        /* Show all directories with more than x% of the changes */
1892        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1893        gather_dirstat(options, &dir, changed, "", 0);
1894}
1895
1896static void free_diffstat_info(struct diffstat_t *diffstat)
1897{
1898        int i;
1899        for (i = 0; i < diffstat->nr; i++) {
1900                struct diffstat_file *f = diffstat->files[i];
1901                if (f->name != f->print_name)
1902                        free(f->print_name);
1903                free(f->name);
1904                free(f->from_name);
1905                free(f);
1906        }
1907        free(diffstat->files);
1908}
1909
1910struct checkdiff_t {
1911        const char *filename;
1912        int lineno;
1913        int conflict_marker_size;
1914        struct diff_options *o;
1915        unsigned ws_rule;
1916        unsigned status;
1917};
1918
1919static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1920{
1921        char firstchar;
1922        int cnt;
1923
1924        if (len < marker_size + 1)
1925                return 0;
1926        firstchar = line[0];
1927        switch (firstchar) {
1928        case '=': case '>': case '<': case '|':
1929                break;
1930        default:
1931                return 0;
1932        }
1933        for (cnt = 1; cnt < marker_size; cnt++)
1934                if (line[cnt] != firstchar)
1935                        return 0;
1936        /* line[1] thru line[marker_size-1] are same as firstchar */
1937        if (len < marker_size + 1 || !isspace(line[marker_size]))
1938                return 0;
1939        return 1;
1940}
1941
1942static void checkdiff_consume(void *priv, char *line, unsigned long len)
1943{
1944        struct checkdiff_t *data = priv;
1945        int marker_size = data->conflict_marker_size;
1946        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
1947        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
1948        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
1949        char *err;
1950        char *line_prefix = "";
1951        struct strbuf *msgbuf;
1952
1953        assert(data->o);
1954        if (data->o->output_prefix) {
1955                msgbuf = data->o->output_prefix(data->o,
1956                        data->o->output_prefix_data);
1957                line_prefix = msgbuf->buf;
1958        }
1959
1960        if (line[0] == '+') {
1961                unsigned bad;
1962                data->lineno++;
1963                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1964                        data->status |= 1;
1965                        fprintf(data->o->file,
1966                                "%s%s:%d: leftover conflict marker\n",
1967                                line_prefix, data->filename, data->lineno);
1968                }
1969                bad = ws_check(line + 1, len - 1, data->ws_rule);
1970                if (!bad)
1971                        return;
1972                data->status |= bad;
1973                err = whitespace_error_string(bad);
1974                fprintf(data->o->file, "%s%s:%d: %s.\n",
1975                        line_prefix, data->filename, data->lineno, err);
1976                free(err);
1977                emit_line(data->o, set, reset, line, 1);
1978                ws_check_emit(line + 1, len - 1, data->ws_rule,
1979                              data->o->file, set, reset, ws);
1980        } else if (line[0] == ' ') {
1981                data->lineno++;
1982        } else if (line[0] == '@') {
1983                char *plus = strchr(line, '+');
1984                if (plus)
1985                        data->lineno = strtol(plus, NULL, 10) - 1;
1986                else
1987                        die("invalid diff");
1988        }
1989}
1990
1991static unsigned char *deflate_it(char *data,
1992                                 unsigned long size,
1993                                 unsigned long *result_size)
1994{
1995        int bound;
1996        unsigned char *deflated;
1997        git_zstream stream;
1998
1999        memset(&stream, 0, sizeof(stream));
2000        git_deflate_init(&stream, zlib_compression_level);
2001        bound = git_deflate_bound(&stream, size);
2002        deflated = xmalloc(bound);
2003        stream.next_out = deflated;
2004        stream.avail_out = bound;
2005
2006        stream.next_in = (unsigned char *)data;
2007        stream.avail_in = size;
2008        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2009                ; /* nothing */
2010        git_deflate_end(&stream);
2011        *result_size = stream.total_out;
2012        return deflated;
2013}
2014
2015static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2016{
2017        void *cp;
2018        void *delta;
2019        void *deflated;
2020        void *data;
2021        unsigned long orig_size;
2022        unsigned long delta_size;
2023        unsigned long deflate_size;
2024        unsigned long data_size;
2025
2026        /* We could do deflated delta, or we could do just deflated two,
2027         * whichever is smaller.
2028         */
2029        delta = NULL;
2030        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2031        if (one->size && two->size) {
2032                delta = diff_delta(one->ptr, one->size,
2033                                   two->ptr, two->size,
2034                                   &delta_size, deflate_size);
2035                if (delta) {
2036                        void *to_free = delta;
2037                        orig_size = delta_size;
2038                        delta = deflate_it(delta, delta_size, &delta_size);
2039                        free(to_free);
2040                }
2041        }
2042
2043        if (delta && delta_size < deflate_size) {
2044                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2045                free(deflated);
2046                data = delta;
2047                data_size = delta_size;
2048        }
2049        else {
2050                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2051                free(delta);
2052                data = deflated;
2053                data_size = deflate_size;
2054        }
2055
2056        /* emit data encoded in base85 */
2057        cp = data;
2058        while (data_size) {
2059                int bytes = (52 < data_size) ? 52 : data_size;
2060                char line[70];
2061                data_size -= bytes;
2062                if (bytes <= 26)
2063                        line[0] = bytes + 'A' - 1;
2064                else
2065                        line[0] = bytes - 26 + 'a' - 1;
2066                encode_85(line + 1, cp, bytes);
2067                cp = (char *) cp + bytes;
2068                fprintf(file, "%s", prefix);
2069                fputs(line, file);
2070                fputc('\n', file);
2071        }
2072        fprintf(file, "%s\n", prefix);
2073        free(data);
2074}
2075
2076static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2077{
2078        fprintf(file, "%sGIT binary patch\n", prefix);
2079        emit_binary_diff_body(file, one, two, prefix);
2080        emit_binary_diff_body(file, two, one, prefix);
2081}
2082
2083int diff_filespec_is_binary(struct diff_filespec *one)
2084{
2085        if (one->is_binary == -1) {
2086                diff_filespec_load_driver(one);
2087                if (one->driver->binary != -1)
2088                        one->is_binary = one->driver->binary;
2089                else {
2090                        if (!one->data && DIFF_FILE_VALID(one))
2091                                diff_populate_filespec(one, 0);
2092                        if (one->data)
2093                                one->is_binary = buffer_is_binary(one->data,
2094                                                one->size);
2095                        if (one->is_binary == -1)
2096                                one->is_binary = 0;
2097                }
2098        }
2099        return one->is_binary;
2100}
2101
2102static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2103{
2104        diff_filespec_load_driver(one);
2105        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2106}
2107
2108void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2109{
2110        if (!options->a_prefix)
2111                options->a_prefix = a;
2112        if (!options->b_prefix)
2113                options->b_prefix = b;
2114}
2115
2116struct userdiff_driver *get_textconv(struct diff_filespec *one)
2117{
2118        if (!DIFF_FILE_VALID(one))
2119                return NULL;
2120
2121        diff_filespec_load_driver(one);
2122        return userdiff_get_textconv(one->driver);
2123}
2124
2125static void builtin_diff(const char *name_a,
2126                         const char *name_b,
2127                         struct diff_filespec *one,
2128                         struct diff_filespec *two,
2129                         const char *xfrm_msg,
2130                         int must_show_header,
2131                         struct diff_options *o,
2132                         int complete_rewrite)
2133{
2134        mmfile_t mf1, mf2;
2135        const char *lbl[2];
2136        char *a_one, *b_two;
2137        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2138        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2139        const char *a_prefix, *b_prefix;
2140        struct userdiff_driver *textconv_one = NULL;
2141        struct userdiff_driver *textconv_two = NULL;
2142        struct strbuf header = STRBUF_INIT;
2143        struct strbuf *msgbuf;
2144        char *line_prefix = "";
2145
2146        if (o->output_prefix) {
2147                msgbuf = o->output_prefix(o, o->output_prefix_data);
2148                line_prefix = msgbuf->buf;
2149        }
2150
2151        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2152                        (!one->mode || S_ISGITLINK(one->mode)) &&
2153                        (!two->mode || S_ISGITLINK(two->mode))) {
2154                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2155                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2156                show_submodule_summary(o->file, one ? one->path : two->path,
2157                                one->sha1, two->sha1, two->dirty_submodule,
2158                                del, add, reset);
2159                return;
2160        }
2161
2162        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2163                textconv_one = get_textconv(one);
2164                textconv_two = get_textconv(two);
2165        }
2166
2167        diff_set_mnemonic_prefix(o, "a/", "b/");
2168        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2169                a_prefix = o->b_prefix;
2170                b_prefix = o->a_prefix;
2171        } else {
2172                a_prefix = o->a_prefix;
2173                b_prefix = o->b_prefix;
2174        }
2175
2176        /* Never use a non-valid filename anywhere if at all possible */
2177        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2178        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2179
2180        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2181        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2182        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2183        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2184        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2185        if (lbl[0][0] == '/') {
2186                /* /dev/null */
2187                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2188                if (xfrm_msg)
2189                        strbuf_addstr(&header, xfrm_msg);
2190                must_show_header = 1;
2191        }
2192        else if (lbl[1][0] == '/') {
2193                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2194                if (xfrm_msg)
2195                        strbuf_addstr(&header, xfrm_msg);
2196                must_show_header = 1;
2197        }
2198        else {
2199                if (one->mode != two->mode) {
2200                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2201                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2202                        must_show_header = 1;
2203                }
2204                if (xfrm_msg)
2205                        strbuf_addstr(&header, xfrm_msg);
2206
2207                /*
2208                 * we do not run diff between different kind
2209                 * of objects.
2210                 */
2211                if ((one->mode ^ two->mode) & S_IFMT)
2212                        goto free_ab_and_return;
2213                if (complete_rewrite &&
2214                    (textconv_one || !diff_filespec_is_binary(one)) &&
2215                    (textconv_two || !diff_filespec_is_binary(two))) {
2216                        fprintf(o->file, "%s", header.buf);
2217                        strbuf_reset(&header);
2218                        emit_rewrite_diff(name_a, name_b, one, two,
2219                                                textconv_one, textconv_two, o);
2220                        o->found_changes = 1;
2221                        goto free_ab_and_return;
2222                }
2223        }
2224
2225        if (o->irreversible_delete && lbl[1][0] == '/') {
2226                fprintf(o->file, "%s", header.buf);
2227                strbuf_reset(&header);
2228                goto free_ab_and_return;
2229        } else if (!DIFF_OPT_TST(o, TEXT) &&
2230            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2231              (!textconv_two && diff_filespec_is_binary(two)) )) {
2232                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2233                        die("unable to read files to diff");
2234                /* Quite common confusing case */
2235                if (mf1.size == mf2.size &&
2236                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2237                        if (must_show_header)
2238                                fprintf(o->file, "%s", header.buf);
2239                        goto free_ab_and_return;
2240                }
2241                fprintf(o->file, "%s", header.buf);
2242                strbuf_reset(&header);
2243                if (DIFF_OPT_TST(o, BINARY))
2244                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2245                else
2246                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2247                                line_prefix, lbl[0], lbl[1]);
2248                o->found_changes = 1;
2249        } else {
2250                /* Crazy xdl interfaces.. */
2251                const char *diffopts = getenv("GIT_DIFF_OPTS");
2252                xpparam_t xpp;
2253                xdemitconf_t xecfg;
2254                struct emit_callback ecbdata;
2255                const struct userdiff_funcname *pe;
2256
2257                if (must_show_header) {
2258                        fprintf(o->file, "%s", header.buf);
2259                        strbuf_reset(&header);
2260                }
2261
2262                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2263                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2264
2265                pe = diff_funcname_pattern(one);
2266                if (!pe)
2267                        pe = diff_funcname_pattern(two);
2268
2269                memset(&xpp, 0, sizeof(xpp));
2270                memset(&xecfg, 0, sizeof(xecfg));
2271                memset(&ecbdata, 0, sizeof(ecbdata));
2272                ecbdata.label_path = lbl;
2273                ecbdata.color_diff = want_color(o->use_color);
2274                ecbdata.found_changesp = &o->found_changes;
2275                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2276                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2277                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2278                ecbdata.opt = o;
2279                ecbdata.header = header.len ? &header : NULL;
2280                xpp.flags = o->xdl_opts;
2281                xecfg.ctxlen = o->context;
2282                xecfg.interhunkctxlen = o->interhunkcontext;
2283                xecfg.flags = XDL_EMIT_FUNCNAMES;
2284                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2285                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2286                if (pe)
2287                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2288                if (!diffopts)
2289                        ;
2290                else if (!prefixcmp(diffopts, "--unified="))
2291                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2292                else if (!prefixcmp(diffopts, "-u"))
2293                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2294                if (o->word_diff)
2295                        init_diff_words_data(&ecbdata, o, one, two);
2296                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2297                              &xpp, &xecfg);
2298                if (o->word_diff)
2299                        free_diff_words_data(&ecbdata);
2300                if (textconv_one)
2301                        free(mf1.ptr);
2302                if (textconv_two)
2303                        free(mf2.ptr);
2304                xdiff_clear_find_func(&xecfg);
2305        }
2306
2307 free_ab_and_return:
2308        strbuf_release(&header);
2309        diff_free_filespec_data(one);
2310        diff_free_filespec_data(two);
2311        free(a_one);
2312        free(b_two);
2313        return;
2314}
2315
2316static void builtin_diffstat(const char *name_a, const char *name_b,
2317                             struct diff_filespec *one,
2318                             struct diff_filespec *two,
2319                             struct diffstat_t *diffstat,
2320                             struct diff_options *o,
2321                             int complete_rewrite)
2322{
2323        mmfile_t mf1, mf2;
2324        struct diffstat_file *data;
2325
2326        data = diffstat_add(diffstat, name_a, name_b);
2327
2328        if (!one || !two) {
2329                data->is_unmerged = 1;
2330                return;
2331        }
2332
2333        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2334                data->is_binary = 1;
2335                data->added = diff_filespec_size(two);
2336                data->deleted = diff_filespec_size(one);
2337        }
2338
2339        else if (complete_rewrite) {
2340                diff_populate_filespec(one, 0);
2341                diff_populate_filespec(two, 0);
2342                data->deleted = count_lines(one->data, one->size);
2343                data->added = count_lines(two->data, two->size);
2344        }
2345
2346        else {
2347                /* Crazy xdl interfaces.. */
2348                xpparam_t xpp;
2349                xdemitconf_t xecfg;
2350
2351                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2352                        die("unable to read files to diff");
2353
2354                memset(&xpp, 0, sizeof(xpp));
2355                memset(&xecfg, 0, sizeof(xecfg));
2356                xpp.flags = o->xdl_opts;
2357                xecfg.ctxlen = o->context;
2358                xecfg.interhunkctxlen = o->interhunkcontext;
2359                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2360                              &xpp, &xecfg);
2361        }
2362
2363        diff_free_filespec_data(one);
2364        diff_free_filespec_data(two);
2365}
2366
2367static void builtin_checkdiff(const char *name_a, const char *name_b,
2368                              const char *attr_path,
2369                              struct diff_filespec *one,
2370                              struct diff_filespec *two,
2371                              struct diff_options *o)
2372{
2373        mmfile_t mf1, mf2;
2374        struct checkdiff_t data;
2375
2376        if (!two)
2377                return;
2378
2379        memset(&data, 0, sizeof(data));
2380        data.filename = name_b ? name_b : name_a;
2381        data.lineno = 0;
2382        data.o = o;
2383        data.ws_rule = whitespace_rule(attr_path);
2384        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2385
2386        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2387                die("unable to read files to diff");
2388
2389        /*
2390         * All the other codepaths check both sides, but not checking
2391         * the "old" side here is deliberate.  We are checking the newly
2392         * introduced changes, and as long as the "new" side is text, we
2393         * can and should check what it introduces.
2394         */
2395        if (diff_filespec_is_binary(two))
2396                goto free_and_return;
2397        else {
2398                /* Crazy xdl interfaces.. */
2399                xpparam_t xpp;
2400                xdemitconf_t xecfg;
2401
2402                memset(&xpp, 0, sizeof(xpp));
2403                memset(&xecfg, 0, sizeof(xecfg));
2404                xecfg.ctxlen = 1; /* at least one context line */
2405                xpp.flags = 0;
2406                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2407                              &xpp, &xecfg);
2408
2409                if (data.ws_rule & WS_BLANK_AT_EOF) {
2410                        struct emit_callback ecbdata;
2411                        int blank_at_eof;
2412
2413                        ecbdata.ws_rule = data.ws_rule;
2414                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2415                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2416
2417                        if (blank_at_eof) {
2418                                static char *err;
2419                                if (!err)
2420                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2421                                fprintf(o->file, "%s:%d: %s.\n",
2422                                        data.filename, blank_at_eof, err);
2423                                data.status = 1; /* report errors */
2424                        }
2425                }
2426        }
2427 free_and_return:
2428        diff_free_filespec_data(one);
2429        diff_free_filespec_data(two);
2430        if (data.status)
2431                DIFF_OPT_SET(o, CHECK_FAILED);
2432}
2433
2434struct diff_filespec *alloc_filespec(const char *path)
2435{
2436        int namelen = strlen(path);
2437        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2438
2439        memset(spec, 0, sizeof(*spec));
2440        spec->path = (char *)(spec + 1);
2441        memcpy(spec->path, path, namelen+1);
2442        spec->count = 1;
2443        spec->is_binary = -1;
2444        return spec;
2445}
2446
2447void free_filespec(struct diff_filespec *spec)
2448{
2449        if (!--spec->count) {
2450                diff_free_filespec_data(spec);
2451                free(spec);
2452        }
2453}
2454
2455void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2456                   unsigned short mode)
2457{
2458        if (mode) {
2459                spec->mode = canon_mode(mode);
2460                hashcpy(spec->sha1, sha1);
2461                spec->sha1_valid = !is_null_sha1(sha1);
2462        }
2463}
2464
2465/*
2466 * Given a name and sha1 pair, if the index tells us the file in
2467 * the work tree has that object contents, return true, so that
2468 * prepare_temp_file() does not have to inflate and extract.
2469 */
2470static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2471{
2472        struct cache_entry *ce;
2473        struct stat st;
2474        int pos, len;
2475
2476        /*
2477         * We do not read the cache ourselves here, because the
2478         * benchmark with my previous version that always reads cache
2479         * shows that it makes things worse for diff-tree comparing
2480         * two linux-2.6 kernel trees in an already checked out work
2481         * tree.  This is because most diff-tree comparisons deal with
2482         * only a small number of files, while reading the cache is
2483         * expensive for a large project, and its cost outweighs the
2484         * savings we get by not inflating the object to a temporary
2485         * file.  Practically, this code only helps when we are used
2486         * by diff-cache --cached, which does read the cache before
2487         * calling us.
2488         */
2489        if (!active_cache)
2490                return 0;
2491
2492        /* We want to avoid the working directory if our caller
2493         * doesn't need the data in a normal file, this system
2494         * is rather slow with its stat/open/mmap/close syscalls,
2495         * and the object is contained in a pack file.  The pack
2496         * is probably already open and will be faster to obtain
2497         * the data through than the working directory.  Loose
2498         * objects however would tend to be slower as they need
2499         * to be individually opened and inflated.
2500         */
2501        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2502                return 0;
2503
2504        len = strlen(name);
2505        pos = cache_name_pos(name, len);
2506        if (pos < 0)
2507                return 0;
2508        ce = active_cache[pos];
2509
2510        /*
2511         * This is not the sha1 we are looking for, or
2512         * unreusable because it is not a regular file.
2513         */
2514        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2515                return 0;
2516
2517        /*
2518         * If ce is marked as "assume unchanged", there is no
2519         * guarantee that work tree matches what we are looking for.
2520         */
2521        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2522                return 0;
2523
2524        /*
2525         * If ce matches the file in the work tree, we can reuse it.
2526         */
2527        if (ce_uptodate(ce) ||
2528            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2529                return 1;
2530
2531        return 0;
2532}
2533
2534static int populate_from_stdin(struct diff_filespec *s)
2535{
2536        struct strbuf buf = STRBUF_INIT;
2537        size_t size = 0;
2538
2539        if (strbuf_read(&buf, 0, 0) < 0)
2540                return error("error while reading from stdin %s",
2541                                     strerror(errno));
2542
2543        s->should_munmap = 0;
2544        s->data = strbuf_detach(&buf, &size);
2545        s->size = size;
2546        s->should_free = 1;
2547        return 0;
2548}
2549
2550static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2551{
2552        int len;
2553        char *data = xmalloc(100), *dirty = "";
2554
2555        /* Are we looking at the work tree? */
2556        if (s->dirty_submodule)
2557                dirty = "-dirty";
2558
2559        len = snprintf(data, 100,
2560                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2561        s->data = data;
2562        s->size = len;
2563        s->should_free = 1;
2564        if (size_only) {
2565                s->data = NULL;
2566                free(data);
2567        }
2568        return 0;
2569}
2570
2571/*
2572 * While doing rename detection and pickaxe operation, we may need to
2573 * grab the data for the blob (or file) for our own in-core comparison.
2574 * diff_filespec has data and size fields for this purpose.
2575 */
2576int diff_populate_filespec(struct diff_filespec *s, int size_only)
2577{
2578        int err = 0;
2579        if (!DIFF_FILE_VALID(s))
2580                die("internal error: asking to populate invalid file.");
2581        if (S_ISDIR(s->mode))
2582                return -1;
2583
2584        if (s->data)
2585                return 0;
2586
2587        if (size_only && 0 < s->size)
2588                return 0;
2589
2590        if (S_ISGITLINK(s->mode))
2591                return diff_populate_gitlink(s, size_only);
2592
2593        if (!s->sha1_valid ||
2594            reuse_worktree_file(s->path, s->sha1, 0)) {
2595                struct strbuf buf = STRBUF_INIT;
2596                struct stat st;
2597                int fd;
2598
2599                if (!strcmp(s->path, "-"))
2600                        return populate_from_stdin(s);
2601
2602                if (lstat(s->path, &st) < 0) {
2603                        if (errno == ENOENT) {
2604                        err_empty:
2605                                err = -1;
2606                        empty:
2607                                s->data = (char *)"";
2608                                s->size = 0;
2609                                return err;
2610                        }
2611                }
2612                s->size = xsize_t(st.st_size);
2613                if (!s->size)
2614                        goto empty;
2615                if (S_ISLNK(st.st_mode)) {
2616                        struct strbuf sb = STRBUF_INIT;
2617
2618                        if (strbuf_readlink(&sb, s->path, s->size))
2619                                goto err_empty;
2620                        s->size = sb.len;
2621                        s->data = strbuf_detach(&sb, NULL);
2622                        s->should_free = 1;
2623                        return 0;
2624                }
2625                if (size_only)
2626                        return 0;
2627                fd = open(s->path, O_RDONLY);
2628                if (fd < 0)
2629                        goto err_empty;
2630                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2631                close(fd);
2632                s->should_munmap = 1;
2633
2634                /*
2635                 * Convert from working tree format to canonical git format
2636                 */
2637                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2638                        size_t size = 0;
2639                        munmap(s->data, s->size);
2640                        s->should_munmap = 0;
2641                        s->data = strbuf_detach(&buf, &size);
2642                        s->size = size;
2643                        s->should_free = 1;
2644                }
2645        }
2646        else {
2647                enum object_type type;
2648                if (size_only) {
2649                        type = sha1_object_info(s->sha1, &s->size);
2650                        if (type < 0)
2651                                die("unable to read %s", sha1_to_hex(s->sha1));
2652                } else {
2653                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2654                        if (!s->data)
2655                                die("unable to read %s", sha1_to_hex(s->sha1));
2656                        s->should_free = 1;
2657                }
2658        }
2659        return 0;
2660}
2661
2662void diff_free_filespec_blob(struct diff_filespec *s)
2663{
2664        if (s->should_free)
2665                free(s->data);
2666        else if (s->should_munmap)
2667                munmap(s->data, s->size);
2668
2669        if (s->should_free || s->should_munmap) {
2670                s->should_free = s->should_munmap = 0;
2671                s->data = NULL;
2672        }
2673}
2674
2675void diff_free_filespec_data(struct diff_filespec *s)
2676{
2677        diff_free_filespec_blob(s);
2678        free(s->cnt_data);
2679        s->cnt_data = NULL;
2680}
2681
2682static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2683                           void *blob,
2684                           unsigned long size,
2685                           const unsigned char *sha1,
2686                           int mode)
2687{
2688        int fd;
2689        struct strbuf buf = STRBUF_INIT;
2690        struct strbuf template = STRBUF_INIT;
2691        char *path_dup = xstrdup(path);
2692        const char *base = basename(path_dup);
2693
2694        /* Generate "XXXXXX_basename.ext" */
2695        strbuf_addstr(&template, "XXXXXX_");
2696        strbuf_addstr(&template, base);
2697
2698        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2699                        strlen(base) + 1);
2700        if (fd < 0)
2701                die_errno("unable to create temp-file");
2702        if (convert_to_working_tree(path,
2703                        (const char *)blob, (size_t)size, &buf)) {
2704                blob = buf.buf;
2705                size = buf.len;
2706        }
2707        if (write_in_full(fd, blob, size) != size)
2708                die_errno("unable to write temp-file");
2709        close(fd);
2710        temp->name = temp->tmp_path;
2711        strcpy(temp->hex, sha1_to_hex(sha1));
2712        temp->hex[40] = 0;
2713        sprintf(temp->mode, "%06o", mode);
2714        strbuf_release(&buf);
2715        strbuf_release(&template);
2716        free(path_dup);
2717}
2718
2719static struct diff_tempfile *prepare_temp_file(const char *name,
2720                struct diff_filespec *one)
2721{
2722        struct diff_tempfile *temp = claim_diff_tempfile();
2723
2724        if (!DIFF_FILE_VALID(one)) {
2725        not_a_valid_file:
2726                /* A '-' entry produces this for file-2, and
2727                 * a '+' entry produces this for file-1.
2728                 */
2729                temp->name = "/dev/null";
2730                strcpy(temp->hex, ".");
2731                strcpy(temp->mode, ".");
2732                return temp;
2733        }
2734
2735        if (!remove_tempfile_installed) {
2736                atexit(remove_tempfile);
2737                sigchain_push_common(remove_tempfile_on_signal);
2738                remove_tempfile_installed = 1;
2739        }
2740
2741        if (!one->sha1_valid ||
2742            reuse_worktree_file(name, one->sha1, 1)) {
2743                struct stat st;
2744                if (lstat(name, &st) < 0) {
2745                        if (errno == ENOENT)
2746                                goto not_a_valid_file;
2747                        die_errno("stat(%s)", name);
2748                }
2749                if (S_ISLNK(st.st_mode)) {
2750                        struct strbuf sb = STRBUF_INIT;
2751                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2752                                die_errno("readlink(%s)", name);
2753                        prep_temp_blob(name, temp, sb.buf, sb.len,
2754                                       (one->sha1_valid ?
2755                                        one->sha1 : null_sha1),
2756                                       (one->sha1_valid ?
2757                                        one->mode : S_IFLNK));
2758                        strbuf_release(&sb);
2759                }
2760                else {
2761                        /* we can borrow from the file in the work tree */
2762                        temp->name = name;
2763                        if (!one->sha1_valid)
2764                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2765                        else
2766                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2767                        /* Even though we may sometimes borrow the
2768                         * contents from the work tree, we always want
2769                         * one->mode.  mode is trustworthy even when
2770                         * !(one->sha1_valid), as long as
2771                         * DIFF_FILE_VALID(one).
2772                         */
2773                        sprintf(temp->mode, "%06o", one->mode);
2774                }
2775                return temp;
2776        }
2777        else {
2778                if (diff_populate_filespec(one, 0))
2779                        die("cannot read data blob for %s", one->path);
2780                prep_temp_blob(name, temp, one->data, one->size,
2781                               one->sha1, one->mode);
2782        }
2783        return temp;
2784}
2785
2786/* An external diff command takes:
2787 *
2788 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2789 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2790 *
2791 */
2792static void run_external_diff(const char *pgm,
2793                              const char *name,
2794                              const char *other,
2795                              struct diff_filespec *one,
2796                              struct diff_filespec *two,
2797                              const char *xfrm_msg,
2798                              int complete_rewrite)
2799{
2800        const char *spawn_arg[10];
2801        int retval;
2802        const char **arg = &spawn_arg[0];
2803
2804        if (one && two) {
2805                struct diff_tempfile *temp_one, *temp_two;
2806                const char *othername = (other ? other : name);
2807                temp_one = prepare_temp_file(name, one);
2808                temp_two = prepare_temp_file(othername, two);
2809                *arg++ = pgm;
2810                *arg++ = name;
2811                *arg++ = temp_one->name;
2812                *arg++ = temp_one->hex;
2813                *arg++ = temp_one->mode;
2814                *arg++ = temp_two->name;
2815                *arg++ = temp_two->hex;
2816                *arg++ = temp_two->mode;
2817                if (other) {
2818                        *arg++ = other;
2819                        *arg++ = xfrm_msg;
2820                }
2821        } else {
2822                *arg++ = pgm;
2823                *arg++ = name;
2824        }
2825        *arg = NULL;
2826        fflush(NULL);
2827        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2828        remove_tempfile();
2829        if (retval) {
2830                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2831                exit(1);
2832        }
2833}
2834
2835static int similarity_index(struct diff_filepair *p)
2836{
2837        return p->score * 100 / MAX_SCORE;
2838}
2839
2840static void fill_metainfo(struct strbuf *msg,
2841                          const char *name,
2842                          const char *other,
2843                          struct diff_filespec *one,
2844                          struct diff_filespec *two,
2845                          struct diff_options *o,
2846                          struct diff_filepair *p,
2847                          int *must_show_header,
2848                          int use_color)
2849{
2850        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2851        const char *reset = diff_get_color(use_color, DIFF_RESET);
2852        struct strbuf *msgbuf;
2853        char *line_prefix = "";
2854
2855        *must_show_header = 1;
2856        if (o->output_prefix) {
2857                msgbuf = o->output_prefix(o, o->output_prefix_data);
2858                line_prefix = msgbuf->buf;
2859        }
2860        strbuf_init(msg, PATH_MAX * 2 + 300);
2861        switch (p->status) {
2862        case DIFF_STATUS_COPIED:
2863                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2864                            line_prefix, set, similarity_index(p));
2865                strbuf_addf(msg, "%s\n%s%scopy from ",
2866                            reset,  line_prefix, set);
2867                quote_c_style(name, msg, NULL, 0);
2868                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2869                quote_c_style(other, msg, NULL, 0);
2870                strbuf_addf(msg, "%s\n", reset);
2871                break;
2872        case DIFF_STATUS_RENAMED:
2873                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2874                            line_prefix, set, similarity_index(p));
2875                strbuf_addf(msg, "%s\n%s%srename from ",
2876                            reset, line_prefix, set);
2877                quote_c_style(name, msg, NULL, 0);
2878                strbuf_addf(msg, "%s\n%s%srename to ",
2879                            reset, line_prefix, set);
2880                quote_c_style(other, msg, NULL, 0);
2881                strbuf_addf(msg, "%s\n", reset);
2882                break;
2883        case DIFF_STATUS_MODIFIED:
2884                if (p->score) {
2885                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2886                                    line_prefix,
2887                                    set, similarity_index(p), reset);
2888                        break;
2889                }
2890                /* fallthru */
2891        default:
2892                *must_show_header = 0;
2893        }
2894        if (one && two && hashcmp(one->sha1, two->sha1)) {
2895                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2896
2897                if (DIFF_OPT_TST(o, BINARY)) {
2898                        mmfile_t mf;
2899                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2900                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2901                                abbrev = 40;
2902                }
2903                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2904                            find_unique_abbrev(one->sha1, abbrev));
2905                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2906                if (one->mode == two->mode)
2907                        strbuf_addf(msg, " %06o", one->mode);
2908                strbuf_addf(msg, "%s\n", reset);
2909        }
2910}
2911
2912static void run_diff_cmd(const char *pgm,
2913                         const char *name,
2914                         const char *other,
2915                         const char *attr_path,
2916                         struct diff_filespec *one,
2917                         struct diff_filespec *two,
2918                         struct strbuf *msg,
2919                         struct diff_options *o,
2920                         struct diff_filepair *p)
2921{
2922        const char *xfrm_msg = NULL;
2923        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2924        int must_show_header = 0;
2925
2926        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2927                pgm = NULL;
2928        else {
2929                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2930                if (drv && drv->external)
2931                        pgm = drv->external;
2932        }
2933
2934        if (msg) {
2935                /*
2936                 * don't use colors when the header is intended for an
2937                 * external diff driver
2938                 */
2939                fill_metainfo(msg, name, other, one, two, o, p,
2940                              &must_show_header,
2941                              want_color(o->use_color) && !pgm);
2942                xfrm_msg = msg->len ? msg->buf : NULL;
2943        }
2944
2945        if (pgm) {
2946                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2947                                  complete_rewrite);
2948                return;
2949        }
2950        if (one && two)
2951                builtin_diff(name, other ? other : name,
2952                             one, two, xfrm_msg, must_show_header,
2953                             o, complete_rewrite);
2954        else
2955                fprintf(o->file, "* Unmerged path %s\n", name);
2956}
2957
2958static void diff_fill_sha1_info(struct diff_filespec *one)
2959{
2960        if (DIFF_FILE_VALID(one)) {
2961                if (!one->sha1_valid) {
2962                        struct stat st;
2963                        if (!strcmp(one->path, "-")) {
2964                                hashcpy(one->sha1, null_sha1);
2965                                return;
2966                        }
2967                        if (lstat(one->path, &st) < 0)
2968                                die_errno("stat '%s'", one->path);
2969                        if (index_path(one->sha1, one->path, &st, 0))
2970                                die("cannot hash %s", one->path);
2971                }
2972        }
2973        else
2974                hashclr(one->sha1);
2975}
2976
2977static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2978{
2979        /* Strip the prefix but do not molest /dev/null and absolute paths */
2980        if (*namep && **namep != '/') {
2981                *namep += prefix_length;
2982                if (**namep == '/')
2983                        ++*namep;
2984        }
2985        if (*otherp && **otherp != '/') {
2986                *otherp += prefix_length;
2987                if (**otherp == '/')
2988                        ++*otherp;
2989        }
2990}
2991
2992static void run_diff(struct diff_filepair *p, struct diff_options *o)
2993{
2994        const char *pgm = external_diff();
2995        struct strbuf msg;
2996        struct diff_filespec *one = p->one;
2997        struct diff_filespec *two = p->two;
2998        const char *name;
2999        const char *other;
3000        const char *attr_path;
3001
3002        name  = p->one->path;
3003        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3004        attr_path = name;
3005        if (o->prefix_length)
3006                strip_prefix(o->prefix_length, &name, &other);
3007
3008        if (DIFF_PAIR_UNMERGED(p)) {
3009                run_diff_cmd(pgm, name, NULL, attr_path,
3010                             NULL, NULL, NULL, o, p);
3011                return;
3012        }
3013
3014        diff_fill_sha1_info(one);
3015        diff_fill_sha1_info(two);
3016
3017        if (!pgm &&
3018            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3019            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3020                /*
3021                 * a filepair that changes between file and symlink
3022                 * needs to be split into deletion and creation.
3023                 */
3024                struct diff_filespec *null = alloc_filespec(two->path);
3025                run_diff_cmd(NULL, name, other, attr_path,
3026                             one, null, &msg, o, p);
3027                free(null);
3028                strbuf_release(&msg);
3029
3030                null = alloc_filespec(one->path);
3031                run_diff_cmd(NULL, name, other, attr_path,
3032                             null, two, &msg, o, p);
3033                free(null);
3034        }
3035        else
3036                run_diff_cmd(pgm, name, other, attr_path,
3037                             one, two, &msg, o, p);
3038
3039        strbuf_release(&msg);
3040}
3041
3042static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3043                         struct diffstat_t *diffstat)
3044{
3045        const char *name;
3046        const char *other;
3047        int complete_rewrite = 0;
3048
3049        if (DIFF_PAIR_UNMERGED(p)) {
3050                /* unmerged */
3051                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3052                return;
3053        }
3054
3055        name = p->one->path;
3056        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3057
3058        if (o->prefix_length)
3059                strip_prefix(o->prefix_length, &name, &other);
3060
3061        diff_fill_sha1_info(p->one);
3062        diff_fill_sha1_info(p->two);
3063
3064        if (p->status == DIFF_STATUS_MODIFIED && p->score)
3065                complete_rewrite = 1;
3066        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3067}
3068
3069static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3070{
3071        const char *name;
3072        const char *other;
3073        const char *attr_path;
3074
3075        if (DIFF_PAIR_UNMERGED(p)) {
3076                /* unmerged */
3077                return;
3078        }
3079
3080        name = p->one->path;
3081        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3082        attr_path = other ? other : name;
3083
3084        if (o->prefix_length)
3085                strip_prefix(o->prefix_length, &name, &other);
3086
3087        diff_fill_sha1_info(p->one);
3088        diff_fill_sha1_info(p->two);
3089
3090        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3091}
3092
3093void diff_setup(struct diff_options *options)
3094{
3095        memcpy(options, &default_diff_options, sizeof(*options));
3096
3097        options->file = stdout;
3098
3099        options->line_termination = '\n';
3100        options->break_opt = -1;
3101        options->rename_limit = -1;
3102        options->dirstat_permille = diff_dirstat_permille_default;
3103        options->context = 3;
3104
3105        options->change = diff_change;
3106        options->add_remove = diff_addremove;
3107        options->use_color = diff_use_color_default;
3108        options->detect_rename = diff_detect_rename_default;
3109
3110        if (diff_no_prefix) {
3111                options->a_prefix = options->b_prefix = "";
3112        } else if (!diff_mnemonic_prefix) {
3113                options->a_prefix = "a/";
3114                options->b_prefix = "b/";
3115        }
3116}
3117
3118int diff_setup_done(struct diff_options *options)
3119{
3120        int count = 0;
3121
3122        if (options->output_format & DIFF_FORMAT_NAME)
3123                count++;
3124        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3125                count++;
3126        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3127                count++;
3128        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3129                count++;
3130        if (count > 1)
3131                die("--name-only, --name-status, --check and -s are mutually exclusive");
3132
3133        /*
3134         * Most of the time we can say "there are changes"
3135         * only by checking if there are changed paths, but
3136         * --ignore-whitespace* options force us to look
3137         * inside contents.
3138         */
3139
3140        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3141            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3142            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3143                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3144        else
3145                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3146
3147        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3148                options->detect_rename = DIFF_DETECT_COPY;
3149
3150        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3151                options->prefix = NULL;
3152        if (options->prefix)
3153                options->prefix_length = strlen(options->prefix);
3154        else
3155                options->prefix_length = 0;
3156
3157        if (options->output_format & (DIFF_FORMAT_NAME |
3158                                      DIFF_FORMAT_NAME_STATUS |
3159                                      DIFF_FORMAT_CHECKDIFF |
3160                                      DIFF_FORMAT_NO_OUTPUT))
3161                options->output_format &= ~(DIFF_FORMAT_RAW |
3162                                            DIFF_FORMAT_NUMSTAT |
3163                                            DIFF_FORMAT_DIFFSTAT |
3164                                            DIFF_FORMAT_SHORTSTAT |
3165                                            DIFF_FORMAT_DIRSTAT |
3166                                            DIFF_FORMAT_SUMMARY |
3167                                            DIFF_FORMAT_PATCH);
3168
3169        /*
3170         * These cases always need recursive; we do not drop caller-supplied
3171         * recursive bits for other formats here.
3172         */
3173        if (options->output_format & (DIFF_FORMAT_PATCH |
3174                                      DIFF_FORMAT_NUMSTAT |
3175                                      DIFF_FORMAT_DIFFSTAT |
3176                                      DIFF_FORMAT_SHORTSTAT |
3177                                      DIFF_FORMAT_DIRSTAT |
3178                                      DIFF_FORMAT_SUMMARY |
3179                                      DIFF_FORMAT_CHECKDIFF))
3180                DIFF_OPT_SET(options, RECURSIVE);
3181        /*
3182         * Also pickaxe would not work very well if you do not say recursive
3183         */
3184        if (options->pickaxe)
3185                DIFF_OPT_SET(options, RECURSIVE);
3186        /*
3187         * When patches are generated, submodules diffed against the work tree
3188         * must be checked for dirtiness too so it can be shown in the output
3189         */
3190        if (options->output_format & DIFF_FORMAT_PATCH)
3191                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3192
3193        if (options->detect_rename && options->rename_limit < 0)
3194                options->rename_limit = diff_rename_limit_default;
3195        if (options->setup & DIFF_SETUP_USE_CACHE) {
3196                if (!active_cache)
3197                        /* read-cache does not die even when it fails
3198                         * so it is safe for us to do this here.  Also
3199                         * it does not smudge active_cache or active_nr
3200                         * when it fails, so we do not have to worry about
3201                         * cleaning it up ourselves either.
3202                         */
3203                        read_cache();
3204        }
3205        if (options->abbrev <= 0 || 40 < options->abbrev)
3206                options->abbrev = 40; /* full */
3207
3208        /*
3209         * It does not make sense to show the first hit we happened
3210         * to have found.  It does not make sense not to return with
3211         * exit code in such a case either.
3212         */
3213        if (DIFF_OPT_TST(options, QUICK)) {
3214                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3215                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3216        }
3217
3218        return 0;
3219}
3220
3221static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3222{
3223        char c, *eq;
3224        int len;
3225
3226        if (*arg != '-')
3227                return 0;
3228        c = *++arg;
3229        if (!c)
3230                return 0;
3231        if (c == arg_short) {
3232                c = *++arg;
3233                if (!c)
3234                        return 1;
3235                if (val && isdigit(c)) {
3236                        char *end;
3237                        int n = strtoul(arg, &end, 10);
3238                        if (*end)
3239                                return 0;
3240                        *val = n;
3241                        return 1;
3242                }
3243                return 0;
3244        }
3245        if (c != '-')
3246                return 0;
3247        arg++;
3248        eq = strchr(arg, '=');
3249        if (eq)
3250                len = eq - arg;
3251        else
3252                len = strlen(arg);
3253        if (!len || strncmp(arg, arg_long, len))
3254                return 0;
3255        if (eq) {
3256                int n;
3257                char *end;
3258                if (!isdigit(*++eq))
3259                        return 0;
3260                n = strtoul(eq, &end, 10);
3261                if (*end)
3262                        return 0;
3263                *val = n;
3264        }
3265        return 1;
3266}
3267
3268static int diff_scoreopt_parse(const char *opt);
3269
3270static inline int short_opt(char opt, const char **argv,
3271                            const char **optarg)
3272{
3273        const char *arg = argv[0];
3274        if (arg[0] != '-' || arg[1] != opt)
3275                return 0;
3276        if (arg[2] != '\0') {
3277                *optarg = arg + 2;
3278                return 1;
3279        }
3280        if (!argv[1])
3281                die("Option '%c' requires a value", opt);
3282        *optarg = argv[1];
3283        return 2;
3284}
3285
3286int parse_long_opt(const char *opt, const char **argv,
3287                   const char **optarg)
3288{
3289        const char *arg = argv[0];
3290        if (arg[0] != '-' || arg[1] != '-')
3291                return 0;
3292        arg += strlen("--");
3293        if (prefixcmp(arg, opt))
3294                return 0;
3295        arg += strlen(opt);
3296        if (*arg == '=') { /* sticked form: --option=value */
3297                *optarg = arg + 1;
3298                return 1;
3299        }
3300        if (*arg != '\0')
3301                return 0;
3302        /* separate form: --option value */
3303        if (!argv[1])
3304                die("Option '--%s' requires a value", opt);
3305        *optarg = argv[1];
3306        return 2;
3307}
3308
3309static int stat_opt(struct diff_options *options, const char **av)
3310{
3311        const char *arg = av[0];
3312        char *end;
3313        int width = options->stat_width;
3314        int name_width = options->stat_name_width;
3315        int count = options->stat_count;
3316        int argcount = 1;
3317
3318        arg += strlen("--stat");
3319        end = (char *)arg;
3320
3321        switch (*arg) {
3322        case '-':
3323                if (!prefixcmp(arg, "-width")) {
3324                        arg += strlen("-width");
3325                        if (*arg == '=')
3326                                width = strtoul(arg + 1, &end, 10);
3327                        else if (!*arg && !av[1])
3328                                die("Option '--stat-width' requires a value");
3329                        else if (!*arg) {
3330                                width = strtoul(av[1], &end, 10);
3331                                argcount = 2;
3332                        }
3333                } else if (!prefixcmp(arg, "-name-width")) {
3334                        arg += strlen("-name-width");
3335                        if (*arg == '=')
3336                                name_width = strtoul(arg + 1, &end, 10);
3337                        else if (!*arg && !av[1])
3338                                die("Option '--stat-name-width' requires a value");
3339                        else if (!*arg) {
3340                                name_width = strtoul(av[1], &end, 10);
3341                                argcount = 2;
3342                        }
3343                } else if (!prefixcmp(arg, "-count")) {
3344                        arg += strlen("-count");
3345                        if (*arg == '=')
3346                                count = strtoul(arg + 1, &end, 10);
3347                        else if (!*arg && !av[1])
3348                                die("Option '--stat-count' requires a value");
3349                        else if (!*arg) {
3350                                count = strtoul(av[1], &end, 10);
3351                                argcount = 2;
3352                        }
3353                }
3354                break;
3355        case '=':
3356                width = strtoul(arg+1, &end, 10);
3357                if (*end == ',')
3358                        name_width = strtoul(end+1, &end, 10);
3359                if (*end == ',')
3360                        count = strtoul(end+1, &end, 10);
3361        }
3362
3363        /* Important! This checks all the error cases! */
3364        if (*end)
3365                return 0;
3366        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3367        options->stat_name_width = name_width;
3368        options->stat_width = width;
3369        options->stat_count = count;
3370        return argcount;
3371}
3372
3373static int parse_dirstat_opt(struct diff_options *options, const char *params)
3374{
3375        struct strbuf errmsg = STRBUF_INIT;
3376        if (parse_dirstat_params(options, params, &errmsg))
3377                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3378                    errmsg.buf);
3379        strbuf_release(&errmsg);
3380        /*
3381         * The caller knows a dirstat-related option is given from the command
3382         * line; allow it to say "return this_function();"
3383         */
3384        options->output_format |= DIFF_FORMAT_DIRSTAT;
3385        return 1;
3386}
3387
3388int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3389{
3390        const char *arg = av[0];
3391        const char *optarg;
3392        int argcount;
3393
3394        /* Output format options */
3395        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3396                options->output_format |= DIFF_FORMAT_PATCH;
3397        else if (opt_arg(arg, 'U', "unified", &options->context))
3398                options->output_format |= DIFF_FORMAT_PATCH;
3399        else if (!strcmp(arg, "--raw"))
3400                options->output_format |= DIFF_FORMAT_RAW;
3401        else if (!strcmp(arg, "--patch-with-raw"))
3402                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3403        else if (!strcmp(arg, "--numstat"))
3404                options->output_format |= DIFF_FORMAT_NUMSTAT;
3405        else if (!strcmp(arg, "--shortstat"))
3406                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3407        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3408                return parse_dirstat_opt(options, "");
3409        else if (!prefixcmp(arg, "-X"))
3410                return parse_dirstat_opt(options, arg + 2);
3411        else if (!prefixcmp(arg, "--dirstat="))
3412                return parse_dirstat_opt(options, arg + 10);
3413        else if (!strcmp(arg, "--cumulative"))
3414                return parse_dirstat_opt(options, "cumulative");
3415        else if (!strcmp(arg, "--dirstat-by-file"))
3416                return parse_dirstat_opt(options, "files");
3417        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3418                parse_dirstat_opt(options, "files");
3419                return parse_dirstat_opt(options, arg + 18);
3420        }
3421        else if (!strcmp(arg, "--check"))
3422                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3423        else if (!strcmp(arg, "--summary"))
3424                options->output_format |= DIFF_FORMAT_SUMMARY;
3425        else if (!strcmp(arg, "--patch-with-stat"))
3426                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3427        else if (!strcmp(arg, "--name-only"))
3428                options->output_format |= DIFF_FORMAT_NAME;
3429        else if (!strcmp(arg, "--name-status"))
3430                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3431        else if (!strcmp(arg, "-s"))
3432                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3433        else if (!prefixcmp(arg, "--stat"))
3434                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3435                return stat_opt(options, av);
3436
3437        /* renames options */
3438        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3439                 !strcmp(arg, "--break-rewrites")) {
3440                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3441                        return error("invalid argument to -B: %s", arg+2);
3442        }
3443        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3444                 !strcmp(arg, "--find-renames")) {
3445                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3446                        return error("invalid argument to -M: %s", arg+2);
3447                options->detect_rename = DIFF_DETECT_RENAME;
3448        }
3449        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3450                options->irreversible_delete = 1;
3451        }
3452        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3453                 !strcmp(arg, "--find-copies")) {
3454                if (options->detect_rename == DIFF_DETECT_COPY)
3455                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3456                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3457                        return error("invalid argument to -C: %s", arg+2);
3458                options->detect_rename = DIFF_DETECT_COPY;
3459        }
3460        else if (!strcmp(arg, "--no-renames"))
3461                options->detect_rename = 0;
3462        else if (!strcmp(arg, "--relative"))
3463                DIFF_OPT_SET(options, RELATIVE_NAME);
3464        else if (!prefixcmp(arg, "--relative=")) {
3465                DIFF_OPT_SET(options, RELATIVE_NAME);
3466                options->prefix = arg + 11;
3467        }
3468
3469        /* xdiff options */
3470        else if (!strcmp(arg, "--minimal"))
3471                DIFF_XDL_SET(options, NEED_MINIMAL);
3472        else if (!strcmp(arg, "--no-minimal"))
3473                DIFF_XDL_CLR(options, NEED_MINIMAL);
3474        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3475                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3476        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3477                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3478        else if (!strcmp(arg, "--ignore-space-at-eol"))
3479                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3480        else if (!strcmp(arg, "--patience"))
3481                DIFF_XDL_SET(options, PATIENCE_DIFF);
3482        else if (!strcmp(arg, "--histogram"))
3483                DIFF_XDL_SET(options, HISTOGRAM_DIFF);
3484
3485        /* flags options */
3486        else if (!strcmp(arg, "--binary")) {
3487                options->output_format |= DIFF_FORMAT_PATCH;
3488                DIFF_OPT_SET(options, BINARY);
3489        }
3490        else if (!strcmp(arg, "--full-index"))
3491                DIFF_OPT_SET(options, FULL_INDEX);
3492        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3493                DIFF_OPT_SET(options, TEXT);
3494        else if (!strcmp(arg, "-R"))
3495                DIFF_OPT_SET(options, REVERSE_DIFF);
3496        else if (!strcmp(arg, "--find-copies-harder"))
3497                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3498        else if (!strcmp(arg, "--follow"))
3499                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3500        else if (!strcmp(arg, "--color"))
3501                options->use_color = 1;
3502        else if (!prefixcmp(arg, "--color=")) {
3503                int value = git_config_colorbool(NULL, arg+8);
3504                if (value < 0)
3505                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3506                options->use_color = value;
3507        }
3508        else if (!strcmp(arg, "--no-color"))
3509                options->use_color = 0;
3510        else if (!strcmp(arg, "--color-words")) {
3511                options->use_color = 1;
3512                options->word_diff = DIFF_WORDS_COLOR;
3513        }
3514        else if (!prefixcmp(arg, "--color-words=")) {
3515                options->use_color = 1;
3516                options->word_diff = DIFF_WORDS_COLOR;
3517                options->word_regex = arg + 14;
3518        }
3519        else if (!strcmp(arg, "--word-diff")) {
3520                if (options->word_diff == DIFF_WORDS_NONE)
3521                        options->word_diff = DIFF_WORDS_PLAIN;
3522        }
3523        else if (!prefixcmp(arg, "--word-diff=")) {
3524                const char *type = arg + 12;
3525                if (!strcmp(type, "plain"))
3526                        options->word_diff = DIFF_WORDS_PLAIN;
3527                else if (!strcmp(type, "color")) {
3528                        options->use_color = 1;
3529                        options->word_diff = DIFF_WORDS_COLOR;
3530                }
3531                else if (!strcmp(type, "porcelain"))
3532                        options->word_diff = DIFF_WORDS_PORCELAIN;
3533                else if (!strcmp(type, "none"))
3534                        options->word_diff = DIFF_WORDS_NONE;
3535                else
3536                        die("bad --word-diff argument: %s", type);
3537        }
3538        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3539                if (options->word_diff == DIFF_WORDS_NONE)
3540                        options->word_diff = DIFF_WORDS_PLAIN;
3541                options->word_regex = optarg;
3542                return argcount;
3543        }
3544        else if (!strcmp(arg, "--exit-code"))
3545                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3546        else if (!strcmp(arg, "--quiet"))
3547                DIFF_OPT_SET(options, QUICK);
3548        else if (!strcmp(arg, "--ext-diff"))
3549                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3550        else if (!strcmp(arg, "--no-ext-diff"))
3551                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3552        else if (!strcmp(arg, "--textconv"))
3553                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3554        else if (!strcmp(arg, "--no-textconv"))
3555                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3556        else if (!strcmp(arg, "--ignore-submodules")) {
3557                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3558                handle_ignore_submodules_arg(options, "all");
3559        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3560                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3561                handle_ignore_submodules_arg(options, arg + 20);
3562        } else if (!strcmp(arg, "--submodule"))
3563                DIFF_OPT_SET(options, SUBMODULE_LOG);
3564        else if (!prefixcmp(arg, "--submodule=")) {
3565                if (!strcmp(arg + 12, "log"))
3566                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3567        }
3568
3569        /* misc options */
3570        else if (!strcmp(arg, "-z"))
3571                options->line_termination = 0;
3572        else if ((argcount = short_opt('l', av, &optarg))) {
3573                options->rename_limit = strtoul(optarg, NULL, 10);
3574                return argcount;
3575        }
3576        else if ((argcount = short_opt('S', av, &optarg))) {
3577                options->pickaxe = optarg;
3578                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3579                return argcount;
3580        } else if ((argcount = short_opt('G', av, &optarg))) {
3581                options->pickaxe = optarg;
3582                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3583                return argcount;
3584        }
3585        else if (!strcmp(arg, "--pickaxe-all"))
3586                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3587        else if (!strcmp(arg, "--pickaxe-regex"))
3588                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3589        else if ((argcount = short_opt('O', av, &optarg))) {
3590                options->orderfile = optarg;
3591                return argcount;
3592        }
3593        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3594                options->filter = optarg;
3595                return argcount;
3596        }
3597        else if (!strcmp(arg, "--abbrev"))
3598                options->abbrev = DEFAULT_ABBREV;
3599        else if (!prefixcmp(arg, "--abbrev=")) {
3600                options->abbrev = strtoul(arg + 9, NULL, 10);
3601                if (options->abbrev < MINIMUM_ABBREV)
3602                        options->abbrev = MINIMUM_ABBREV;
3603                else if (40 < options->abbrev)
3604                        options->abbrev = 40;
3605        }
3606        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3607                options->a_prefix = optarg;
3608                return argcount;
3609        }
3610        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3611                options->b_prefix = optarg;
3612                return argcount;
3613        }
3614        else if (!strcmp(arg, "--no-prefix"))
3615                options->a_prefix = options->b_prefix = "";
3616        else if (opt_arg(arg, '\0', "inter-hunk-context",
3617                         &options->interhunkcontext))
3618                ;
3619        else if (!strcmp(arg, "-W"))
3620                DIFF_OPT_SET(options, FUNCCONTEXT);
3621        else if (!strcmp(arg, "--function-context"))
3622                DIFF_OPT_SET(options, FUNCCONTEXT);
3623        else if (!strcmp(arg, "--no-function-context"))
3624                DIFF_OPT_CLR(options, FUNCCONTEXT);
3625        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3626                options->file = fopen(optarg, "w");
3627                if (!options->file)
3628                        die_errno("Could not open '%s'", optarg);
3629                options->close_file = 1;
3630                return argcount;
3631        } else
3632                return 0;
3633        return 1;
3634}
3635
3636int parse_rename_score(const char **cp_p)
3637{
3638        unsigned long num, scale;
3639        int ch, dot;
3640        const char *cp = *cp_p;
3641
3642        num = 0;
3643        scale = 1;
3644        dot = 0;
3645        for (;;) {
3646                ch = *cp;
3647                if ( !dot && ch == '.' ) {
3648                        scale = 1;
3649                        dot = 1;
3650                } else if ( ch == '%' ) {
3651                        scale = dot ? scale*100 : 100;
3652                        cp++;   /* % is always at the end */
3653                        break;
3654                } else if ( ch >= '0' && ch <= '9' ) {
3655                        if ( scale < 100000 ) {
3656                                scale *= 10;
3657                                num = (num*10) + (ch-'0');
3658                        }
3659                } else {
3660                        break;
3661                }
3662                cp++;
3663        }
3664        *cp_p = cp;
3665
3666        /* user says num divided by scale and we say internally that
3667         * is MAX_SCORE * num / scale.
3668         */
3669        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3670}
3671
3672static int diff_scoreopt_parse(const char *opt)
3673{
3674        int opt1, opt2, cmd;
3675
3676        if (*opt++ != '-')
3677                return -1;
3678        cmd = *opt++;
3679        if (cmd == '-') {
3680                /* convert the long-form arguments into short-form versions */
3681                if (!prefixcmp(opt, "break-rewrites")) {
3682                        opt += strlen("break-rewrites");
3683                        if (*opt == 0 || *opt++ == '=')
3684                                cmd = 'B';
3685                } else if (!prefixcmp(opt, "find-copies")) {
3686                        opt += strlen("find-copies");
3687                        if (*opt == 0 || *opt++ == '=')
3688                                cmd = 'C';
3689                } else if (!prefixcmp(opt, "find-renames")) {
3690                        opt += strlen("find-renames");
3691                        if (*opt == 0 || *opt++ == '=')
3692                                cmd = 'M';
3693                }
3694        }
3695        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3696                return -1; /* that is not a -M, -C nor -B option */
3697
3698        opt1 = parse_rename_score(&opt);
3699        if (cmd != 'B')
3700                opt2 = 0;
3701        else {
3702                if (*opt == 0)
3703                        opt2 = 0;
3704                else if (*opt != '/')
3705                        return -1; /* we expect -B80/99 or -B80 */
3706                else {
3707                        opt++;
3708                        opt2 = parse_rename_score(&opt);
3709                }
3710        }
3711        if (*opt != 0)
3712                return -1;
3713        return opt1 | (opt2 << 16);
3714}
3715
3716struct diff_queue_struct diff_queued_diff;
3717
3718void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3719{
3720        if (queue->alloc <= queue->nr) {
3721                queue->alloc = alloc_nr(queue->alloc);
3722                queue->queue = xrealloc(queue->queue,
3723                                        sizeof(dp) * queue->alloc);
3724        }
3725        queue->queue[queue->nr++] = dp;
3726}
3727
3728struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3729                                 struct diff_filespec *one,
3730                                 struct diff_filespec *two)
3731{
3732        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3733        dp->one = one;
3734        dp->two = two;
3735        if (queue)
3736                diff_q(queue, dp);
3737        return dp;
3738}
3739
3740void diff_free_filepair(struct diff_filepair *p)
3741{
3742        free_filespec(p->one);
3743        free_filespec(p->two);
3744        free(p);
3745}
3746
3747/* This is different from find_unique_abbrev() in that
3748 * it stuffs the result with dots for alignment.
3749 */
3750const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3751{
3752        int abblen;
3753        const char *abbrev;
3754        if (len == 40)
3755                return sha1_to_hex(sha1);
3756
3757        abbrev = find_unique_abbrev(sha1, len);
3758        abblen = strlen(abbrev);
3759        if (abblen < 37) {
3760                static char hex[41];
3761                if (len < abblen && abblen <= len + 2)
3762                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3763                else
3764                        sprintf(hex, "%s...", abbrev);
3765                return hex;
3766        }
3767        return sha1_to_hex(sha1);
3768}
3769
3770static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3771{
3772        int line_termination = opt->line_termination;
3773        int inter_name_termination = line_termination ? '\t' : '\0';
3774        if (opt->output_prefix) {
3775                struct strbuf *msg = NULL;
3776                msg = opt->output_prefix(opt, opt->output_prefix_data);
3777                fprintf(opt->file, "%s", msg->buf);
3778        }
3779
3780        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3781                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3782                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3783                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3784        }
3785        if (p->score) {
3786                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3787                        inter_name_termination);
3788        } else {
3789                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3790        }
3791
3792        if (p->status == DIFF_STATUS_COPIED ||
3793            p->status == DIFF_STATUS_RENAMED) {
3794                const char *name_a, *name_b;
3795                name_a = p->one->path;
3796                name_b = p->two->path;
3797                strip_prefix(opt->prefix_length, &name_a, &name_b);
3798                write_name_quoted(name_a, opt->file, inter_name_termination);
3799                write_name_quoted(name_b, opt->file, line_termination);
3800        } else {
3801                const char *name_a, *name_b;
3802                name_a = p->one->mode ? p->one->path : p->two->path;
3803                name_b = NULL;
3804                strip_prefix(opt->prefix_length, &name_a, &name_b);
3805                write_name_quoted(name_a, opt->file, line_termination);
3806        }
3807}
3808
3809int diff_unmodified_pair(struct diff_filepair *p)
3810{
3811        /* This function is written stricter than necessary to support
3812         * the currently implemented transformers, but the idea is to
3813         * let transformers to produce diff_filepairs any way they want,
3814         * and filter and clean them up here before producing the output.
3815         */
3816        struct diff_filespec *one = p->one, *two = p->two;
3817
3818        if (DIFF_PAIR_UNMERGED(p))
3819                return 0; /* unmerged is interesting */
3820
3821        /* deletion, addition, mode or type change
3822         * and rename are all interesting.
3823         */
3824        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3825            DIFF_PAIR_MODE_CHANGED(p) ||
3826            strcmp(one->path, two->path))
3827                return 0;
3828
3829        /* both are valid and point at the same path.  that is, we are
3830         * dealing with a change.
3831         */
3832        if (one->sha1_valid && two->sha1_valid &&
3833            !hashcmp(one->sha1, two->sha1) &&
3834            !one->dirty_submodule && !two->dirty_submodule)
3835                return 1; /* no change */
3836        if (!one->sha1_valid && !two->sha1_valid)
3837                return 1; /* both look at the same file on the filesystem. */
3838        return 0;
3839}
3840
3841static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3842{
3843        if (diff_unmodified_pair(p))
3844                return;
3845
3846        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3847            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3848                return; /* no tree diffs in patch format */
3849
3850        run_diff(p, o);
3851}
3852
3853static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3854                            struct diffstat_t *diffstat)
3855{
3856        if (diff_unmodified_pair(p))
3857                return;
3858
3859        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3860            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3861                return; /* no useful stat for tree diffs */
3862
3863        run_diffstat(p, o, diffstat);
3864}
3865
3866static void diff_flush_checkdiff(struct diff_filepair *p,
3867                struct diff_options *o)
3868{
3869        if (diff_unmodified_pair(p))
3870                return;
3871
3872        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3873            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3874                return; /* nothing to check in tree diffs */
3875
3876        run_checkdiff(p, o);
3877}
3878
3879int diff_queue_is_empty(void)
3880{
3881        struct diff_queue_struct *q = &diff_queued_diff;
3882        int i;
3883        for (i = 0; i < q->nr; i++)
3884                if (!diff_unmodified_pair(q->queue[i]))
3885                        return 0;
3886        return 1;
3887}
3888
3889#if DIFF_DEBUG
3890void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3891{
3892        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3893                x, one ? one : "",
3894                s->path,
3895                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3896                s->mode,
3897                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3898        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3899                x, one ? one : "",
3900                s->size, s->xfrm_flags);
3901}
3902
3903void diff_debug_filepair(const struct diff_filepair *p, int i)
3904{
3905        diff_debug_filespec(p->one, i, "one");
3906        diff_debug_filespec(p->two, i, "two");
3907        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3908                p->score, p->status ? p->status : '?',
3909                p->one->rename_used, p->broken_pair);
3910}
3911
3912void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3913{
3914        int i;
3915        if (msg)
3916                fprintf(stderr, "%s\n", msg);
3917        fprintf(stderr, "q->nr = %d\n", q->nr);
3918        for (i = 0; i < q->nr; i++) {
3919                struct diff_filepair *p = q->queue[i];
3920                diff_debug_filepair(p, i);
3921        }
3922}
3923#endif
3924
3925static void diff_resolve_rename_copy(void)
3926{
3927        int i;
3928        struct diff_filepair *p;
3929        struct diff_queue_struct *q = &diff_queued_diff;
3930
3931        diff_debug_queue("resolve-rename-copy", q);
3932
3933        for (i = 0; i < q->nr; i++) {
3934                p = q->queue[i];
3935                p->status = 0; /* undecided */
3936                if (DIFF_PAIR_UNMERGED(p))
3937                        p->status = DIFF_STATUS_UNMERGED;
3938                else if (!DIFF_FILE_VALID(p->one))
3939                        p->status = DIFF_STATUS_ADDED;
3940                else if (!DIFF_FILE_VALID(p->two))
3941                        p->status = DIFF_STATUS_DELETED;
3942                else if (DIFF_PAIR_TYPE_CHANGED(p))
3943                        p->status = DIFF_STATUS_TYPE_CHANGED;
3944
3945                /* from this point on, we are dealing with a pair
3946                 * whose both sides are valid and of the same type, i.e.
3947                 * either in-place edit or rename/copy edit.
3948                 */
3949                else if (DIFF_PAIR_RENAME(p)) {
3950                        /*
3951                         * A rename might have re-connected a broken
3952                         * pair up, causing the pathnames to be the
3953                         * same again. If so, that's not a rename at
3954                         * all, just a modification..
3955                         *
3956                         * Otherwise, see if this source was used for
3957                         * multiple renames, in which case we decrement
3958                         * the count, and call it a copy.
3959                         */
3960                        if (!strcmp(p->one->path, p->two->path))
3961                                p->status = DIFF_STATUS_MODIFIED;
3962                        else if (--p->one->rename_used > 0)
3963                                p->status = DIFF_STATUS_COPIED;
3964                        else
3965                                p->status = DIFF_STATUS_RENAMED;
3966                }
3967                else if (hashcmp(p->one->sha1, p->two->sha1) ||
3968                         p->one->mode != p->two->mode ||
3969                         p->one->dirty_submodule ||
3970                         p->two->dirty_submodule ||
3971                         is_null_sha1(p->one->sha1))
3972                        p->status = DIFF_STATUS_MODIFIED;
3973                else {
3974                        /* This is a "no-change" entry and should not
3975                         * happen anymore, but prepare for broken callers.
3976                         */
3977                        error("feeding unmodified %s to diffcore",
3978                              p->one->path);
3979                        p->status = DIFF_STATUS_UNKNOWN;
3980                }
3981        }
3982        diff_debug_queue("resolve-rename-copy done", q);
3983}
3984
3985static int check_pair_status(struct diff_filepair *p)
3986{
3987        switch (p->status) {
3988        case DIFF_STATUS_UNKNOWN:
3989                return 0;
3990        case 0:
3991                die("internal error in diff-resolve-rename-copy");
3992        default:
3993                return 1;
3994        }
3995}
3996
3997static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3998{
3999        int fmt = opt->output_format;
4000
4001        if (fmt & DIFF_FORMAT_CHECKDIFF)
4002                diff_flush_checkdiff(p, opt);
4003        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4004                diff_flush_raw(p, opt);
4005        else if (fmt & DIFF_FORMAT_NAME) {
4006                const char *name_a, *name_b;
4007                name_a = p->two->path;
4008                name_b = NULL;
4009                strip_prefix(opt->prefix_length, &name_a, &name_b);
4010                write_name_quoted(name_a, opt->file, opt->line_termination);
4011        }
4012}
4013
4014static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4015{
4016        if (fs->mode)
4017                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4018        else
4019                fprintf(file, " %s ", newdelete);
4020        write_name_quoted(fs->path, file, '\n');
4021}
4022
4023
4024static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4025                const char *line_prefix)
4026{
4027        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4028                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4029                        p->two->mode, show_name ? ' ' : '\n');
4030                if (show_name) {
4031                        write_name_quoted(p->two->path, file, '\n');
4032                }
4033        }
4034}
4035
4036static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4037                        const char *line_prefix)
4038{
4039        char *names = pprint_rename(p->one->path, p->two->path);
4040
4041        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4042        free(names);
4043        show_mode_change(file, p, 0, line_prefix);
4044}
4045
4046static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4047{
4048        FILE *file = opt->file;
4049        char *line_prefix = "";
4050
4051        if (opt->output_prefix) {
4052                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4053                line_prefix = buf->buf;
4054        }
4055
4056        switch(p->status) {
4057        case DIFF_STATUS_DELETED:
4058                fputs(line_prefix, file);
4059                show_file_mode_name(file, "delete", p->one);
4060                break;
4061        case DIFF_STATUS_ADDED:
4062                fputs(line_prefix, file);
4063                show_file_mode_name(file, "create", p->two);
4064                break;
4065        case DIFF_STATUS_COPIED:
4066                fputs(line_prefix, file);
4067                show_rename_copy(file, "copy", p, line_prefix);
4068                break;
4069        case DIFF_STATUS_RENAMED:
4070                fputs(line_prefix, file);
4071                show_rename_copy(file, "rename", p, line_prefix);
4072                break;
4073        default:
4074                if (p->score) {
4075                        fprintf(file, "%s rewrite ", line_prefix);
4076                        write_name_quoted(p->two->path, file, ' ');
4077                        fprintf(file, "(%d%%)\n", similarity_index(p));
4078                }
4079                show_mode_change(file, p, !p->score, line_prefix);
4080                break;
4081        }
4082}
4083
4084struct patch_id_t {
4085        git_SHA_CTX *ctx;
4086        int patchlen;
4087};
4088
4089static int remove_space(char *line, int len)
4090{
4091        int i;
4092        char *dst = line;
4093        unsigned char c;
4094
4095        for (i = 0; i < len; i++)
4096                if (!isspace((c = line[i])))
4097                        *dst++ = c;
4098
4099        return dst - line;
4100}
4101
4102static void patch_id_consume(void *priv, char *line, unsigned long len)
4103{
4104        struct patch_id_t *data = priv;
4105        int new_len;
4106
4107        /* Ignore line numbers when computing the SHA1 of the patch */
4108        if (!prefixcmp(line, "@@ -"))
4109                return;
4110
4111        new_len = remove_space(line, len);
4112
4113        git_SHA1_Update(data->ctx, line, new_len);
4114        data->patchlen += new_len;
4115}
4116
4117/* returns 0 upon success, and writes result into sha1 */
4118static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4119{
4120        struct diff_queue_struct *q = &diff_queued_diff;
4121        int i;
4122        git_SHA_CTX ctx;
4123        struct patch_id_t data;
4124        char buffer[PATH_MAX * 4 + 20];
4125
4126        git_SHA1_Init(&ctx);
4127        memset(&data, 0, sizeof(struct patch_id_t));
4128        data.ctx = &ctx;
4129
4130        for (i = 0; i < q->nr; i++) {
4131                xpparam_t xpp;
4132                xdemitconf_t xecfg;
4133                mmfile_t mf1, mf2;
4134                struct diff_filepair *p = q->queue[i];
4135                int len1, len2;
4136
4137                memset(&xpp, 0, sizeof(xpp));
4138                memset(&xecfg, 0, sizeof(xecfg));
4139                if (p->status == 0)
4140                        return error("internal diff status error");
4141                if (p->status == DIFF_STATUS_UNKNOWN)
4142                        continue;
4143                if (diff_unmodified_pair(p))
4144                        continue;
4145                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4146                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4147                        continue;
4148                if (DIFF_PAIR_UNMERGED(p))
4149                        continue;
4150
4151                diff_fill_sha1_info(p->one);
4152                diff_fill_sha1_info(p->two);
4153                if (fill_mmfile(&mf1, p->one) < 0 ||
4154                                fill_mmfile(&mf2, p->two) < 0)
4155                        return error("unable to read files to diff");
4156
4157                len1 = remove_space(p->one->path, strlen(p->one->path));
4158                len2 = remove_space(p->two->path, strlen(p->two->path));
4159                if (p->one->mode == 0)
4160                        len1 = snprintf(buffer, sizeof(buffer),
4161                                        "diff--gita/%.*sb/%.*s"
4162                                        "newfilemode%06o"
4163                                        "---/dev/null"
4164                                        "+++b/%.*s",
4165                                        len1, p->one->path,
4166                                        len2, p->two->path,
4167                                        p->two->mode,
4168                                        len2, p->two->path);
4169                else if (p->two->mode == 0)
4170                        len1 = snprintf(buffer, sizeof(buffer),
4171                                        "diff--gita/%.*sb/%.*s"
4172                                        "deletedfilemode%06o"
4173                                        "---a/%.*s"
4174                                        "+++/dev/null",
4175                                        len1, p->one->path,
4176                                        len2, p->two->path,
4177                                        p->one->mode,
4178                                        len1, p->one->path);
4179                else
4180                        len1 = snprintf(buffer, sizeof(buffer),
4181                                        "diff--gita/%.*sb/%.*s"
4182                                        "---a/%.*s"
4183                                        "+++b/%.*s",
4184                                        len1, p->one->path,
4185                                        len2, p->two->path,
4186                                        len1, p->one->path,
4187                                        len2, p->two->path);
4188                git_SHA1_Update(&ctx, buffer, len1);
4189
4190                if (diff_filespec_is_binary(p->one) ||
4191                    diff_filespec_is_binary(p->two)) {
4192                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4193                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4194                        continue;
4195                }
4196
4197                xpp.flags = 0;
4198                xecfg.ctxlen = 3;
4199                xecfg.flags = 0;
4200                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4201                              &xpp, &xecfg);
4202        }
4203
4204        git_SHA1_Final(sha1, &ctx);
4205        return 0;
4206}
4207
4208int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4209{
4210        struct diff_queue_struct *q = &diff_queued_diff;
4211        int i;
4212        int result = diff_get_patch_id(options, sha1);
4213
4214        for (i = 0; i < q->nr; i++)
4215                diff_free_filepair(q->queue[i]);
4216
4217        free(q->queue);
4218        DIFF_QUEUE_CLEAR(q);
4219
4220        return result;
4221}
4222
4223static int is_summary_empty(const struct diff_queue_struct *q)
4224{
4225        int i;
4226
4227        for (i = 0; i < q->nr; i++) {
4228                const struct diff_filepair *p = q->queue[i];
4229
4230                switch (p->status) {
4231                case DIFF_STATUS_DELETED:
4232                case DIFF_STATUS_ADDED:
4233                case DIFF_STATUS_COPIED:
4234                case DIFF_STATUS_RENAMED:
4235                        return 0;
4236                default:
4237                        if (p->score)
4238                                return 0;
4239                        if (p->one->mode && p->two->mode &&
4240                            p->one->mode != p->two->mode)
4241                                return 0;
4242                        break;
4243                }
4244        }
4245        return 1;
4246}
4247
4248static const char rename_limit_warning[] =
4249"inexact rename detection was skipped due to too many files.";
4250
4251static const char degrade_cc_to_c_warning[] =
4252"only found copies from modified paths due to too many files.";
4253
4254static const char rename_limit_advice[] =
4255"you may want to set your %s variable to at least "
4256"%d and retry the command.";
4257
4258void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4259{
4260        if (degraded_cc)
4261                warning(degrade_cc_to_c_warning);
4262        else if (needed)
4263                warning(rename_limit_warning);
4264        else
4265                return;
4266        if (0 < needed && needed < 32767)
4267                warning(rename_limit_advice, varname, needed);
4268}
4269
4270void diff_flush(struct diff_options *options)
4271{
4272        struct diff_queue_struct *q = &diff_queued_diff;
4273        int i, output_format = options->output_format;
4274        int separator = 0;
4275        int dirstat_by_line = 0;
4276
4277        /*
4278         * Order: raw, stat, summary, patch
4279         * or:    name/name-status/checkdiff (other bits clear)
4280         */
4281        if (!q->nr)
4282                goto free_queue;
4283
4284        if (output_format & (DIFF_FORMAT_RAW |
4285                             DIFF_FORMAT_NAME |
4286                             DIFF_FORMAT_NAME_STATUS |
4287                             DIFF_FORMAT_CHECKDIFF)) {
4288                for (i = 0; i < q->nr; i++) {
4289                        struct diff_filepair *p = q->queue[i];
4290                        if (check_pair_status(p))
4291                                flush_one_pair(p, options);
4292                }
4293                separator++;
4294        }
4295
4296        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4297                dirstat_by_line = 1;
4298
4299        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4300            dirstat_by_line) {
4301                struct diffstat_t diffstat;
4302
4303                memset(&diffstat, 0, sizeof(struct diffstat_t));
4304                for (i = 0; i < q->nr; i++) {
4305                        struct diff_filepair *p = q->queue[i];
4306                        if (check_pair_status(p))
4307                                diff_flush_stat(p, options, &diffstat);
4308                }
4309                if (output_format & DIFF_FORMAT_NUMSTAT)
4310                        show_numstat(&diffstat, options);
4311                if (output_format & DIFF_FORMAT_DIFFSTAT)
4312                        show_stats(&diffstat, options);
4313                if (output_format & DIFF_FORMAT_SHORTSTAT)
4314                        show_shortstats(&diffstat, options);
4315                if (output_format & DIFF_FORMAT_DIRSTAT)
4316                        show_dirstat_by_line(&diffstat, options);
4317                free_diffstat_info(&diffstat);
4318                separator++;
4319        }
4320        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4321                show_dirstat(options);
4322
4323        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4324                for (i = 0; i < q->nr; i++) {
4325                        diff_summary(options, q->queue[i]);
4326                }
4327                separator++;
4328        }
4329
4330        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4331            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4332            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4333                /*
4334                 * run diff_flush_patch for the exit status. setting
4335                 * options->file to /dev/null should be safe, becaue we
4336                 * aren't supposed to produce any output anyway.
4337                 */
4338                if (options->close_file)
4339                        fclose(options->file);
4340                options->file = fopen("/dev/null", "w");
4341                if (!options->file)
4342                        die_errno("Could not open /dev/null");
4343                options->close_file = 1;
4344                for (i = 0; i < q->nr; i++) {
4345                        struct diff_filepair *p = q->queue[i];
4346                        if (check_pair_status(p))
4347                                diff_flush_patch(p, options);
4348                        if (options->found_changes)
4349                                break;
4350                }
4351        }
4352
4353        if (output_format & DIFF_FORMAT_PATCH) {
4354                if (separator) {
4355                        putc(options->line_termination, options->file);
4356                        if (options->stat_sep) {
4357                                /* attach patch instead of inline */
4358                                fputs(options->stat_sep, options->file);
4359                        }
4360                }
4361
4362                for (i = 0; i < q->nr; i++) {
4363                        struct diff_filepair *p = q->queue[i];
4364                        if (check_pair_status(p))
4365                                diff_flush_patch(p, options);
4366                }
4367        }
4368
4369        if (output_format & DIFF_FORMAT_CALLBACK)
4370                options->format_callback(q, options, options->format_callback_data);
4371
4372        for (i = 0; i < q->nr; i++)
4373                diff_free_filepair(q->queue[i]);
4374free_queue:
4375        free(q->queue);
4376        DIFF_QUEUE_CLEAR(q);
4377        if (options->close_file)
4378                fclose(options->file);
4379
4380        /*
4381         * Report the content-level differences with HAS_CHANGES;
4382         * diff_addremove/diff_change does not set the bit when
4383         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4384         */
4385        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4386                if (options->found_changes)
4387                        DIFF_OPT_SET(options, HAS_CHANGES);
4388                else
4389                        DIFF_OPT_CLR(options, HAS_CHANGES);
4390        }
4391}
4392
4393static void diffcore_apply_filter(const char *filter)
4394{
4395        int i;
4396        struct diff_queue_struct *q = &diff_queued_diff;
4397        struct diff_queue_struct outq;
4398        DIFF_QUEUE_CLEAR(&outq);
4399
4400        if (!filter)
4401                return;
4402
4403        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4404                int found;
4405                for (i = found = 0; !found && i < q->nr; i++) {
4406                        struct diff_filepair *p = q->queue[i];
4407                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4408                             ((p->score &&
4409                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4410                              (!p->score &&
4411                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4412                            ((p->status != DIFF_STATUS_MODIFIED) &&
4413                             strchr(filter, p->status)))
4414                                found++;
4415                }
4416                if (found)
4417                        return;
4418
4419                /* otherwise we will clear the whole queue
4420                 * by copying the empty outq at the end of this
4421                 * function, but first clear the current entries
4422                 * in the queue.
4423                 */
4424                for (i = 0; i < q->nr; i++)
4425                        diff_free_filepair(q->queue[i]);
4426        }
4427        else {
4428                /* Only the matching ones */
4429                for (i = 0; i < q->nr; i++) {
4430                        struct diff_filepair *p = q->queue[i];
4431
4432                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4433                             ((p->score &&
4434                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4435                              (!p->score &&
4436                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4437                            ((p->status != DIFF_STATUS_MODIFIED) &&
4438                             strchr(filter, p->status)))
4439                                diff_q(&outq, p);
4440                        else
4441                                diff_free_filepair(p);
4442                }
4443        }
4444        free(q->queue);
4445        *q = outq;
4446}
4447
4448/* Check whether two filespecs with the same mode and size are identical */
4449static int diff_filespec_is_identical(struct diff_filespec *one,
4450                                      struct diff_filespec *two)
4451{
4452        if (S_ISGITLINK(one->mode))
4453                return 0;
4454        if (diff_populate_filespec(one, 0))
4455                return 0;
4456        if (diff_populate_filespec(two, 0))
4457                return 0;
4458        return !memcmp(one->data, two->data, one->size);
4459}
4460
4461static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4462{
4463        int i;
4464        struct diff_queue_struct *q = &diff_queued_diff;
4465        struct diff_queue_struct outq;
4466        DIFF_QUEUE_CLEAR(&outq);
4467
4468        for (i = 0; i < q->nr; i++) {
4469                struct diff_filepair *p = q->queue[i];
4470
4471                /*
4472                 * 1. Entries that come from stat info dirtiness
4473                 *    always have both sides (iow, not create/delete),
4474                 *    one side of the object name is unknown, with
4475                 *    the same mode and size.  Keep the ones that
4476                 *    do not match these criteria.  They have real
4477                 *    differences.
4478                 *
4479                 * 2. At this point, the file is known to be modified,
4480                 *    with the same mode and size, and the object
4481                 *    name of one side is unknown.  Need to inspect
4482                 *    the identical contents.
4483                 */
4484                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4485                    !DIFF_FILE_VALID(p->two) ||
4486                    (p->one->sha1_valid && p->two->sha1_valid) ||
4487                    (p->one->mode != p->two->mode) ||
4488                    diff_populate_filespec(p->one, 1) ||
4489                    diff_populate_filespec(p->two, 1) ||
4490                    (p->one->size != p->two->size) ||
4491                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4492                        diff_q(&outq, p);
4493                else {
4494                        /*
4495                         * The caller can subtract 1 from skip_stat_unmatch
4496                         * to determine how many paths were dirty only
4497                         * due to stat info mismatch.
4498                         */
4499                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4500                                diffopt->skip_stat_unmatch++;
4501                        diff_free_filepair(p);
4502                }
4503        }
4504        free(q->queue);
4505        *q = outq;
4506}
4507
4508static int diffnamecmp(const void *a_, const void *b_)
4509{
4510        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4511        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4512        const char *name_a, *name_b;
4513
4514        name_a = a->one ? a->one->path : a->two->path;
4515        name_b = b->one ? b->one->path : b->two->path;
4516        return strcmp(name_a, name_b);
4517}
4518
4519void diffcore_fix_diff_index(struct diff_options *options)
4520{
4521        struct diff_queue_struct *q = &diff_queued_diff;
4522        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4523}
4524
4525void diffcore_std(struct diff_options *options)
4526{
4527        if (options->skip_stat_unmatch)
4528                diffcore_skip_stat_unmatch(options);
4529        if (!options->found_follow) {
4530                /* See try_to_follow_renames() in tree-diff.c */
4531                if (options->break_opt != -1)
4532                        diffcore_break(options->break_opt);
4533                if (options->detect_rename)
4534                        diffcore_rename(options);
4535                if (options->break_opt != -1)
4536                        diffcore_merge_broken();
4537        }
4538        if (options->pickaxe)
4539                diffcore_pickaxe(options);
4540        if (options->orderfile)
4541                diffcore_order(options->orderfile);
4542        if (!options->found_follow)
4543                /* See try_to_follow_renames() in tree-diff.c */
4544                diff_resolve_rename_copy();
4545        diffcore_apply_filter(options->filter);
4546
4547        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4548                DIFF_OPT_SET(options, HAS_CHANGES);
4549        else
4550                DIFF_OPT_CLR(options, HAS_CHANGES);
4551
4552        options->found_follow = 0;
4553}
4554
4555int diff_result_code(struct diff_options *opt, int status)
4556{
4557        int result = 0;
4558
4559        diff_warn_rename_limit("diff.renamelimit",
4560                               opt->needed_rename_limit,
4561                               opt->degraded_cc_to_c);
4562        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4563            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4564                return status;
4565        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4566            DIFF_OPT_TST(opt, HAS_CHANGES))
4567                result |= 01;
4568        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4569            DIFF_OPT_TST(opt, CHECK_FAILED))
4570                result |= 02;
4571        return result;
4572}
4573
4574int diff_can_quit_early(struct diff_options *opt)
4575{
4576        return (DIFF_OPT_TST(opt, QUICK) &&
4577                !opt->filter &&
4578                DIFF_OPT_TST(opt, HAS_CHANGES));
4579}
4580
4581/*
4582 * Shall changes to this submodule be ignored?
4583 *
4584 * Submodule changes can be configured to be ignored separately for each path,
4585 * but that configuration can be overridden from the command line.
4586 */
4587static int is_submodule_ignored(const char *path, struct diff_options *options)
4588{
4589        int ignored = 0;
4590        unsigned orig_flags = options->flags;
4591        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4592                set_diffopt_flags_from_submodule_config(options, path);
4593        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4594                ignored = 1;
4595        options->flags = orig_flags;
4596        return ignored;
4597}
4598
4599void diff_addremove(struct diff_options *options,
4600                    int addremove, unsigned mode,
4601                    const unsigned char *sha1,
4602                    const char *concatpath, unsigned dirty_submodule)
4603{
4604        struct diff_filespec *one, *two;
4605
4606        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4607                return;
4608
4609        /* This may look odd, but it is a preparation for
4610         * feeding "there are unchanged files which should
4611         * not produce diffs, but when you are doing copy
4612         * detection you would need them, so here they are"
4613         * entries to the diff-core.  They will be prefixed
4614         * with something like '=' or '*' (I haven't decided
4615         * which but should not make any difference).
4616         * Feeding the same new and old to diff_change()
4617         * also has the same effect.
4618         * Before the final output happens, they are pruned after
4619         * merged into rename/copy pairs as appropriate.
4620         */
4621        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4622                addremove = (addremove == '+' ? '-' :
4623                             addremove == '-' ? '+' : addremove);
4624
4625        if (options->prefix &&
4626            strncmp(concatpath, options->prefix, options->prefix_length))
4627                return;
4628
4629        one = alloc_filespec(concatpath);
4630        two = alloc_filespec(concatpath);
4631
4632        if (addremove != '+')
4633                fill_filespec(one, sha1, mode);
4634        if (addremove != '-') {
4635                fill_filespec(two, sha1, mode);
4636                two->dirty_submodule = dirty_submodule;
4637        }
4638
4639        diff_queue(&diff_queued_diff, one, two);
4640        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4641                DIFF_OPT_SET(options, HAS_CHANGES);
4642}
4643
4644void diff_change(struct diff_options *options,
4645                 unsigned old_mode, unsigned new_mode,
4646                 const unsigned char *old_sha1,
4647                 const unsigned char *new_sha1,
4648                 const char *concatpath,
4649                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4650{
4651        struct diff_filespec *one, *two;
4652
4653        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4654            is_submodule_ignored(concatpath, options))
4655                return;
4656
4657        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4658                unsigned tmp;
4659                const unsigned char *tmp_c;
4660                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4661                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4662                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4663                        new_dirty_submodule = tmp;
4664        }
4665
4666        if (options->prefix &&
4667            strncmp(concatpath, options->prefix, options->prefix_length))
4668                return;
4669
4670        one = alloc_filespec(concatpath);
4671        two = alloc_filespec(concatpath);
4672        fill_filespec(one, old_sha1, old_mode);
4673        fill_filespec(two, new_sha1, new_mode);
4674        one->dirty_submodule = old_dirty_submodule;
4675        two->dirty_submodule = new_dirty_submodule;
4676
4677        diff_queue(&diff_queued_diff, one, two);
4678        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4679                DIFF_OPT_SET(options, HAS_CHANGES);
4680}
4681
4682struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4683{
4684        struct diff_filepair *pair;
4685        struct diff_filespec *one, *two;
4686
4687        if (options->prefix &&
4688            strncmp(path, options->prefix, options->prefix_length))
4689                return NULL;
4690
4691        one = alloc_filespec(path);
4692        two = alloc_filespec(path);
4693        pair = diff_queue(&diff_queued_diff, one, two);
4694        pair->is_unmerged = 1;
4695        return pair;
4696}
4697
4698static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4699                size_t *outsize)
4700{
4701        struct diff_tempfile *temp;
4702        const char *argv[3];
4703        const char **arg = argv;
4704        struct child_process child;
4705        struct strbuf buf = STRBUF_INIT;
4706        int err = 0;
4707
4708        temp = prepare_temp_file(spec->path, spec);
4709        *arg++ = pgm;
4710        *arg++ = temp->name;
4711        *arg = NULL;
4712
4713        memset(&child, 0, sizeof(child));
4714        child.use_shell = 1;
4715        child.argv = argv;
4716        child.out = -1;
4717        if (start_command(&child)) {
4718                remove_tempfile();
4719                return NULL;
4720        }
4721
4722        if (strbuf_read(&buf, child.out, 0) < 0)
4723                err = error("error reading from textconv command '%s'", pgm);
4724        close(child.out);
4725
4726        if (finish_command(&child) || err) {
4727                strbuf_release(&buf);
4728                remove_tempfile();
4729                return NULL;
4730        }
4731        remove_tempfile();
4732
4733        return strbuf_detach(&buf, outsize);
4734}
4735
4736size_t fill_textconv(struct userdiff_driver *driver,
4737                     struct diff_filespec *df,
4738                     char **outbuf)
4739{
4740        size_t size;
4741
4742        if (!driver || !driver->textconv) {
4743                if (!DIFF_FILE_VALID(df)) {
4744                        *outbuf = "";
4745                        return 0;
4746                }
4747                if (diff_populate_filespec(df, 0))
4748                        die("unable to read files to diff");
4749                *outbuf = df->data;
4750                return df->size;
4751        }
4752
4753        if (driver->textconv_cache && df->sha1_valid) {
4754                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4755                                          &size);
4756                if (*outbuf)
4757                        return size;
4758        }
4759
4760        *outbuf = run_textconv(driver->textconv, df, &size);
4761        if (!*outbuf)
4762                die("unable to read files to diff");
4763
4764        if (driver->textconv_cache && df->sha1_valid) {
4765                /* ignore errors, as we might be in a readonly repository */
4766                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4767                                size);
4768                /*
4769                 * we could save up changes and flush them all at the end,
4770                 * but we would need an extra call after all diffing is done.
4771                 * Since generating a cache entry is the slow path anyway,
4772                 * this extra overhead probably isn't a big deal.
4773                 */
4774                notes_cache_write(driver->textconv_cache);
4775        }
4776
4777        return size;
4778}