diff.con commit Make --color-words work well with --graph (4297c0a)
   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 = 200;
  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;
  33
  34static char diff_colors[][COLOR_MAXLEN] = {
  35        GIT_COLOR_RESET,
  36        GIT_COLOR_NORMAL,       /* PLAIN */
  37        GIT_COLOR_BOLD,         /* METAINFO */
  38        GIT_COLOR_CYAN,         /* FRAGINFO */
  39        GIT_COLOR_RED,          /* OLD */
  40        GIT_COLOR_GREEN,        /* NEW */
  41        GIT_COLOR_YELLOW,       /* COMMIT */
  42        GIT_COLOR_BG_RED,       /* WHITESPACE */
  43        GIT_COLOR_NORMAL,       /* FUNCINFO */
  44};
  45
  46static void diff_filespec_load_driver(struct diff_filespec *one);
  47static size_t fill_textconv(struct userdiff_driver *driver,
  48                            struct diff_filespec *df, char **outbuf);
  49
  50static int parse_diff_color_slot(const char *var, int ofs)
  51{
  52        if (!strcasecmp(var+ofs, "plain"))
  53                return DIFF_PLAIN;
  54        if (!strcasecmp(var+ofs, "meta"))
  55                return DIFF_METAINFO;
  56        if (!strcasecmp(var+ofs, "frag"))
  57                return DIFF_FRAGINFO;
  58        if (!strcasecmp(var+ofs, "old"))
  59                return DIFF_FILE_OLD;
  60        if (!strcasecmp(var+ofs, "new"))
  61                return DIFF_FILE_NEW;
  62        if (!strcasecmp(var+ofs, "commit"))
  63                return DIFF_COMMIT;
  64        if (!strcasecmp(var+ofs, "whitespace"))
  65                return DIFF_WHITESPACE;
  66        if (!strcasecmp(var+ofs, "func"))
  67                return DIFF_FUNCINFO;
  68        return -1;
  69}
  70
  71static int git_config_rename(const char *var, const char *value)
  72{
  73        if (!value)
  74                return DIFF_DETECT_RENAME;
  75        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
  76                return  DIFF_DETECT_COPY;
  77        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
  78}
  79
  80/*
  81 * These are to give UI layer defaults.
  82 * The core-level commands such as git-diff-files should
  83 * never be affected by the setting of diff.renames
  84 * the user happens to have in the configuration file.
  85 */
  86int git_diff_ui_config(const char *var, const char *value, void *cb)
  87{
  88        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
  89                diff_use_color_default = git_config_colorbool(var, value, -1);
  90                return 0;
  91        }
  92        if (!strcmp(var, "diff.renames")) {
  93                diff_detect_rename_default = git_config_rename(var, value);
  94                return 0;
  95        }
  96        if (!strcmp(var, "diff.autorefreshindex")) {
  97                diff_auto_refresh_index = git_config_bool(var, value);
  98                return 0;
  99        }
 100        if (!strcmp(var, "diff.mnemonicprefix")) {
 101                diff_mnemonic_prefix = git_config_bool(var, value);
 102                return 0;
 103        }
 104        if (!strcmp(var, "diff.external"))
 105                return git_config_string(&external_diff_cmd_cfg, var, value);
 106        if (!strcmp(var, "diff.wordregex"))
 107                return git_config_string(&diff_word_regex_cfg, var, value);
 108
 109        return git_diff_basic_config(var, value, cb);
 110}
 111
 112int git_diff_basic_config(const char *var, const char *value, void *cb)
 113{
 114        if (!strcmp(var, "diff.renamelimit")) {
 115                diff_rename_limit_default = git_config_int(var, value);
 116                return 0;
 117        }
 118
 119        switch (userdiff_config(var, value)) {
 120                case 0: break;
 121                case -1: return -1;
 122                default: return 0;
 123        }
 124
 125        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 126                int slot = parse_diff_color_slot(var, 11);
 127                if (slot < 0)
 128                        return 0;
 129                if (!value)
 130                        return config_error_nonbool(var);
 131                color_parse(value, var, diff_colors[slot]);
 132                return 0;
 133        }
 134
 135        /* like GNU diff's --suppress-blank-empty option  */
 136        if (!strcmp(var, "diff.suppressblankempty") ||
 137                        /* for backwards compatibility */
 138                        !strcmp(var, "diff.suppress-blank-empty")) {
 139                diff_suppress_blank_empty = git_config_bool(var, value);
 140                return 0;
 141        }
 142
 143        return git_color_default_config(var, value, cb);
 144}
 145
 146static char *quote_two(const char *one, const char *two)
 147{
 148        int need_one = quote_c_style(one, NULL, NULL, 1);
 149        int need_two = quote_c_style(two, NULL, NULL, 1);
 150        struct strbuf res = STRBUF_INIT;
 151
 152        if (need_one + need_two) {
 153                strbuf_addch(&res, '"');
 154                quote_c_style(one, &res, NULL, 1);
 155                quote_c_style(two, &res, NULL, 1);
 156                strbuf_addch(&res, '"');
 157        } else {
 158                strbuf_addstr(&res, one);
 159                strbuf_addstr(&res, two);
 160        }
 161        return strbuf_detach(&res, NULL);
 162}
 163
 164static const char *external_diff(void)
 165{
 166        static const char *external_diff_cmd = NULL;
 167        static int done_preparing = 0;
 168
 169        if (done_preparing)
 170                return external_diff_cmd;
 171        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 172        if (!external_diff_cmd)
 173                external_diff_cmd = external_diff_cmd_cfg;
 174        done_preparing = 1;
 175        return external_diff_cmd;
 176}
 177
 178static struct diff_tempfile {
 179        const char *name; /* filename external diff should read from */
 180        char hex[41];
 181        char mode[10];
 182        char tmp_path[PATH_MAX];
 183} diff_temp[2];
 184
 185typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 186
 187struct emit_callback {
 188        int color_diff;
 189        unsigned ws_rule;
 190        int blank_at_eof_in_preimage;
 191        int blank_at_eof_in_postimage;
 192        int lno_in_preimage;
 193        int lno_in_postimage;
 194        sane_truncate_fn truncate;
 195        const char **label_path;
 196        struct diff_words_data *diff_words;
 197        struct diff_options *opt;
 198        int *found_changesp;
 199        struct strbuf *header;
 200};
 201
 202static int count_lines(const char *data, int size)
 203{
 204        int count, ch, completely_empty = 1, nl_just_seen = 0;
 205        count = 0;
 206        while (0 < size--) {
 207                ch = *data++;
 208                if (ch == '\n') {
 209                        count++;
 210                        nl_just_seen = 1;
 211                        completely_empty = 0;
 212                }
 213                else {
 214                        nl_just_seen = 0;
 215                        completely_empty = 0;
 216                }
 217        }
 218        if (completely_empty)
 219                return 0;
 220        if (!nl_just_seen)
 221                count++; /* no trailing newline */
 222        return count;
 223}
 224
 225static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 226{
 227        if (!DIFF_FILE_VALID(one)) {
 228                mf->ptr = (char *)""; /* does not matter */
 229                mf->size = 0;
 230                return 0;
 231        }
 232        else if (diff_populate_filespec(one, 0))
 233                return -1;
 234
 235        mf->ptr = one->data;
 236        mf->size = one->size;
 237        return 0;
 238}
 239
 240static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 241{
 242        char *ptr = mf->ptr;
 243        long size = mf->size;
 244        int cnt = 0;
 245
 246        if (!size)
 247                return cnt;
 248        ptr += size - 1; /* pointing at the very end */
 249        if (*ptr != '\n')
 250                ; /* incomplete line */
 251        else
 252                ptr--; /* skip the last LF */
 253        while (mf->ptr < ptr) {
 254                char *prev_eol;
 255                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 256                        if (*prev_eol == '\n')
 257                                break;
 258                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 259                        break;
 260                cnt++;
 261                ptr = prev_eol - 1;
 262        }
 263        return cnt;
 264}
 265
 266static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 267                               struct emit_callback *ecbdata)
 268{
 269        int l1, l2, at;
 270        unsigned ws_rule = ecbdata->ws_rule;
 271        l1 = count_trailing_blank(mf1, ws_rule);
 272        l2 = count_trailing_blank(mf2, ws_rule);
 273        if (l2 <= l1) {
 274                ecbdata->blank_at_eof_in_preimage = 0;
 275                ecbdata->blank_at_eof_in_postimage = 0;
 276                return;
 277        }
 278        at = count_lines(mf1->ptr, mf1->size);
 279        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 280
 281        at = count_lines(mf2->ptr, mf2->size);
 282        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 283}
 284
 285static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 286                        int first, const char *line, int len)
 287{
 288        int has_trailing_newline, has_trailing_carriage_return;
 289        int nofirst;
 290        FILE *file = o->file;
 291
 292        if (o->output_prefix) {
 293                struct strbuf *msg = NULL;
 294                msg = o->output_prefix(o, o->output_prefix_data);
 295                assert(msg);
 296                fwrite(msg->buf, msg->len, 1, file);
 297        }
 298
 299        if (len == 0) {
 300                has_trailing_newline = (first == '\n');
 301                has_trailing_carriage_return = (!has_trailing_newline &&
 302                                                (first == '\r'));
 303                nofirst = has_trailing_newline || has_trailing_carriage_return;
 304        } else {
 305                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 306                if (has_trailing_newline)
 307                        len--;
 308                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 309                if (has_trailing_carriage_return)
 310                        len--;
 311                nofirst = 0;
 312        }
 313
 314        if (len || !nofirst) {
 315                fputs(set, file);
 316                if (!nofirst)
 317                        fputc(first, file);
 318                fwrite(line, len, 1, file);
 319                fputs(reset, file);
 320        }
 321        if (has_trailing_carriage_return)
 322                fputc('\r', file);
 323        if (has_trailing_newline)
 324                fputc('\n', file);
 325}
 326
 327static void emit_line(struct diff_options *o, const char *set, const char *reset,
 328                      const char *line, int len)
 329{
 330        emit_line_0(o, set, reset, line[0], line+1, len-1);
 331}
 332
 333static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 334{
 335        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 336              ecbdata->blank_at_eof_in_preimage &&
 337              ecbdata->blank_at_eof_in_postimage &&
 338              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 339              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 340                return 0;
 341        return ws_blank_line(line, len, ecbdata->ws_rule);
 342}
 343
 344static void emit_add_line(const char *reset,
 345                          struct emit_callback *ecbdata,
 346                          const char *line, int len)
 347{
 348        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 349        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 350
 351        if (!*ws)
 352                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 353        else if (new_blank_line_at_eof(ecbdata, line, len))
 354                /* Blank line at EOF - paint '+' as well */
 355                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 356        else {
 357                /* Emit just the prefix, then the rest. */
 358                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 359                ws_check_emit(line, len, ecbdata->ws_rule,
 360                              ecbdata->opt->file, set, reset, ws);
 361        }
 362}
 363
 364static void emit_hunk_header(struct emit_callback *ecbdata,
 365                             const char *line, int len)
 366{
 367        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 368        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 369        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 370        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 371        static const char atat[2] = { '@', '@' };
 372        const char *cp, *ep;
 373        struct strbuf msgbuf = STRBUF_INIT;
 374        int org_len = len;
 375        int i = 1;
 376
 377        /*
 378         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 379         * it always is at least 10 bytes long.
 380         */
 381        if (len < 10 ||
 382            memcmp(line, atat, 2) ||
 383            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 384                emit_line(ecbdata->opt, plain, reset, line, len);
 385                return;
 386        }
 387        ep += 2; /* skip over @@ */
 388
 389        /* The hunk header in fraginfo color */
 390        strbuf_add(&msgbuf, frag, strlen(frag));
 391        strbuf_add(&msgbuf, line, ep - line);
 392        strbuf_add(&msgbuf, reset, strlen(reset));
 393
 394        /*
 395         * trailing "\r\n"
 396         */
 397        for ( ; i < 3; i++)
 398                if (line[len - i] == '\r' || line[len - i] == '\n')
 399                        len--;
 400
 401        /* blank before the func header */
 402        for (cp = ep; ep - line < len; ep++)
 403                if (*ep != ' ' && *ep != '\t')
 404                        break;
 405        if (ep != cp) {
 406                strbuf_add(&msgbuf, plain, strlen(plain));
 407                strbuf_add(&msgbuf, cp, ep - cp);
 408                strbuf_add(&msgbuf, reset, strlen(reset));
 409        }
 410
 411        if (ep < line + len) {
 412                strbuf_add(&msgbuf, func, strlen(func));
 413                strbuf_add(&msgbuf, ep, line + len - ep);
 414                strbuf_add(&msgbuf, reset, strlen(reset));
 415        }
 416
 417        strbuf_add(&msgbuf, line + len, org_len - len);
 418        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 419        strbuf_release(&msgbuf);
 420}
 421
 422static struct diff_tempfile *claim_diff_tempfile(void) {
 423        int i;
 424        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 425                if (!diff_temp[i].name)
 426                        return diff_temp + i;
 427        die("BUG: diff is failing to clean up its tempfiles");
 428}
 429
 430static int remove_tempfile_installed;
 431
 432static void remove_tempfile(void)
 433{
 434        int i;
 435        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 436                if (diff_temp[i].name == diff_temp[i].tmp_path)
 437                        unlink_or_warn(diff_temp[i].name);
 438                diff_temp[i].name = NULL;
 439        }
 440}
 441
 442static void remove_tempfile_on_signal(int signo)
 443{
 444        remove_tempfile();
 445        sigchain_pop(signo);
 446        raise(signo);
 447}
 448
 449static void print_line_count(FILE *file, int count)
 450{
 451        switch (count) {
 452        case 0:
 453                fprintf(file, "0,0");
 454                break;
 455        case 1:
 456                fprintf(file, "1");
 457                break;
 458        default:
 459                fprintf(file, "1,%d", count);
 460                break;
 461        }
 462}
 463
 464static void emit_rewrite_lines(struct emit_callback *ecb,
 465                               int prefix, const char *data, int size)
 466{
 467        const char *endp = NULL;
 468        static const char *nneof = " No newline at end of file\n";
 469        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 470        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 471
 472        while (0 < size) {
 473                int len;
 474
 475                endp = memchr(data, '\n', size);
 476                len = endp ? (endp - data + 1) : size;
 477                if (prefix != '+') {
 478                        ecb->lno_in_preimage++;
 479                        emit_line_0(ecb->opt, old, reset, '-',
 480                                    data, len);
 481                } else {
 482                        ecb->lno_in_postimage++;
 483                        emit_add_line(reset, ecb, data, len);
 484                }
 485                size -= len;
 486                data += len;
 487        }
 488        if (!endp) {
 489                const char *plain = diff_get_color(ecb->color_diff,
 490                                                   DIFF_PLAIN);
 491                emit_line_0(ecb->opt, plain, reset, '\\',
 492                            nneof, strlen(nneof));
 493        }
 494}
 495
 496static void emit_rewrite_diff(const char *name_a,
 497                              const char *name_b,
 498                              struct diff_filespec *one,
 499                              struct diff_filespec *two,
 500                              struct userdiff_driver *textconv_one,
 501                              struct userdiff_driver *textconv_two,
 502                              struct diff_options *o)
 503{
 504        int lc_a, lc_b;
 505        int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
 506        const char *name_a_tab, *name_b_tab;
 507        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 508        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 509        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 510        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 511        const char *a_prefix, *b_prefix;
 512        char *data_one, *data_two;
 513        size_t size_one, size_two;
 514        struct emit_callback ecbdata;
 515        char *line_prefix = "";
 516        struct strbuf *msgbuf;
 517
 518        if (o && o->output_prefix) {
 519                msgbuf = o->output_prefix(o, o->output_prefix_data);
 520                line_prefix = msgbuf->buf;
 521        }
 522
 523        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 524                a_prefix = o->b_prefix;
 525                b_prefix = o->a_prefix;
 526        } else {
 527                a_prefix = o->a_prefix;
 528                b_prefix = o->b_prefix;
 529        }
 530
 531        name_a += (*name_a == '/');
 532        name_b += (*name_b == '/');
 533        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 534        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 535
 536        strbuf_reset(&a_name);
 537        strbuf_reset(&b_name);
 538        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 539        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 540
 541        size_one = fill_textconv(textconv_one, one, &data_one);
 542        size_two = fill_textconv(textconv_two, two, &data_two);
 543
 544        memset(&ecbdata, 0, sizeof(ecbdata));
 545        ecbdata.color_diff = color_diff;
 546        ecbdata.found_changesp = &o->found_changes;
 547        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 548        ecbdata.opt = o;
 549        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 550                mmfile_t mf1, mf2;
 551                mf1.ptr = (char *)data_one;
 552                mf2.ptr = (char *)data_two;
 553                mf1.size = size_one;
 554                mf2.size = size_two;
 555                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 556        }
 557        ecbdata.lno_in_preimage = 1;
 558        ecbdata.lno_in_postimage = 1;
 559
 560        lc_a = count_lines(data_one, size_one);
 561        lc_b = count_lines(data_two, size_two);
 562        fprintf(o->file,
 563                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 564                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 565                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 566                line_prefix, fraginfo);
 567        print_line_count(o->file, lc_a);
 568        fprintf(o->file, " +");
 569        print_line_count(o->file, lc_b);
 570        fprintf(o->file, " @@%s\n", reset);
 571        if (lc_a)
 572                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 573        if (lc_b)
 574                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 575        if (textconv_one)
 576                free((char *)data_one);
 577        if (textconv_two)
 578                free((char *)data_two);
 579}
 580
 581struct diff_words_buffer {
 582        mmfile_t text;
 583        long alloc;
 584        struct diff_words_orig {
 585                const char *begin, *end;
 586        } *orig;
 587        int orig_nr, orig_alloc;
 588};
 589
 590static void diff_words_append(char *line, unsigned long len,
 591                struct diff_words_buffer *buffer)
 592{
 593        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 594        line++;
 595        len--;
 596        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 597        buffer->text.size += len;
 598        buffer->text.ptr[buffer->text.size] = '\0';
 599}
 600
 601struct diff_words_style_elem
 602{
 603        const char *prefix;
 604        const char *suffix;
 605        const char *color; /* NULL; filled in by the setup code if
 606                            * color is enabled */
 607};
 608
 609struct diff_words_style
 610{
 611        enum diff_words_type type;
 612        struct diff_words_style_elem new, old, ctx;
 613        const char *newline;
 614};
 615
 616struct diff_words_style diff_words_styles[] = {
 617        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 618        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 619        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 620};
 621
 622struct diff_words_data {
 623        struct diff_words_buffer minus, plus;
 624        const char *current_plus;
 625        int last_minus;
 626        struct diff_options *opt;
 627        regex_t *word_regex;
 628        enum diff_words_type type;
 629        struct diff_words_style *style;
 630};
 631
 632static int fn_out_diff_words_write_helper(FILE *fp,
 633                                          struct diff_words_style_elem *st_el,
 634                                          const char *newline,
 635                                          size_t count, const char *buf,
 636                                          const char *line_prefix)
 637{
 638        int print = 0;
 639
 640        while (count) {
 641                char *p = memchr(buf, '\n', count);
 642                if (print)
 643                        fputs(line_prefix, fp);
 644                if (p != buf) {
 645                        if (st_el->color && fputs(st_el->color, fp) < 0)
 646                                return -1;
 647                        if (fputs(st_el->prefix, fp) < 0 ||
 648                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 649                            fputs(st_el->suffix, fp) < 0)
 650                                return -1;
 651                        if (st_el->color && *st_el->color
 652                            && fputs(GIT_COLOR_RESET, fp) < 0)
 653                                return -1;
 654                }
 655                if (!p)
 656                        return 0;
 657                if (fputs(newline, fp) < 0)
 658                        return -1;
 659                count -= p + 1 - buf;
 660                buf = p + 1;
 661                print = 1;
 662        }
 663        return 0;
 664}
 665
 666/*
 667 * '--color-words' algorithm can be described as:
 668 *
 669 *   1. collect a the minus/plus lines of a diff hunk, divided into
 670 *      minus-lines and plus-lines;
 671 *
 672 *   2. break both minus-lines and plus-lines into words and
 673 *      place them into two mmfile_t with one word for each line;
 674 *
 675 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 676 *
 677 * And for the common parts of the both file, we output the plus side text.
 678 * diff_words->current_plus is used to trace the current position of the plus file
 679 * which printed. diff_words->last_minus is used to trace the last minus word
 680 * printed.
 681 *
 682 * For '--graph' to work with '--color-words', we need to output the graph prefix
 683 * on each line of color words output. Generally, there are two conditions on
 684 * which we should output the prefix.
 685 *
 686 *   1. diff_words->last_minus == 0 &&
 687 *      diff_words->current_plus == diff_words->plus.text.ptr
 688 *
 689 *      that is: the plus text must start as a new line, and if there is no minus
 690 *      word printed, a graph prefix must be printed.
 691 *
 692 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 693 *      *(diff_words->current_plus - 1) == '\n'
 694 *
 695 *      that is: a graph prefix must be printed following a '\n'
 696 */
 697static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 698{
 699        if ((diff_words->last_minus == 0 &&
 700                diff_words->current_plus == diff_words->plus.text.ptr) ||
 701                (diff_words->current_plus > diff_words->plus.text.ptr &&
 702                *(diff_words->current_plus - 1) == '\n')) {
 703                return 1;
 704        } else {
 705                return 0;
 706        }
 707}
 708
 709static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 710{
 711        struct diff_words_data *diff_words = priv;
 712        struct diff_words_style *style = diff_words->style;
 713        int minus_first, minus_len, plus_first, plus_len;
 714        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 715        struct diff_options *opt = diff_words->opt;
 716        struct strbuf *msgbuf;
 717        char *line_prefix = "";
 718
 719        if (line[0] != '@' || parse_hunk_header(line, len,
 720                        &minus_first, &minus_len, &plus_first, &plus_len))
 721                return;
 722
 723        assert(opt);
 724        if (opt->output_prefix) {
 725                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 726                line_prefix = msgbuf->buf;
 727        }
 728
 729        /* POSIX requires that first be decremented by one if len == 0... */
 730        if (minus_len) {
 731                minus_begin = diff_words->minus.orig[minus_first].begin;
 732                minus_end =
 733                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 734        } else
 735                minus_begin = minus_end =
 736                        diff_words->minus.orig[minus_first].end;
 737
 738        if (plus_len) {
 739                plus_begin = diff_words->plus.orig[plus_first].begin;
 740                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 741        } else
 742                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 743
 744        if (color_words_output_graph_prefix(diff_words)) {
 745                fputs(line_prefix, diff_words->opt->file);
 746        }
 747        if (diff_words->current_plus != plus_begin) {
 748                fn_out_diff_words_write_helper(diff_words->opt->file,
 749                                &style->ctx, style->newline,
 750                                plus_begin - diff_words->current_plus,
 751                                diff_words->current_plus, line_prefix);
 752                if (*(plus_begin - 1) == '\n')
 753                        fputs(line_prefix, diff_words->opt->file);
 754        }
 755        if (minus_begin != minus_end) {
 756                fn_out_diff_words_write_helper(diff_words->opt->file,
 757                                &style->old, style->newline,
 758                                minus_end - minus_begin, minus_begin,
 759                                line_prefix);
 760        }
 761        if (plus_begin != plus_end) {
 762                fn_out_diff_words_write_helper(diff_words->opt->file,
 763                                &style->new, style->newline,
 764                                plus_end - plus_begin, plus_begin,
 765                                line_prefix);
 766        }
 767
 768        diff_words->current_plus = plus_end;
 769        diff_words->last_minus = minus_first;
 770}
 771
 772/* This function starts looking at *begin, and returns 0 iff a word was found. */
 773static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 774                int *begin, int *end)
 775{
 776        if (word_regex && *begin < buffer->size) {
 777                regmatch_t match[1];
 778                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 779                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 780                                        '\n', match[0].rm_eo - match[0].rm_so);
 781                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 782                        *begin += match[0].rm_so;
 783                        return *begin >= *end;
 784                }
 785                return -1;
 786        }
 787
 788        /* find the next word */
 789        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 790                (*begin)++;
 791        if (*begin >= buffer->size)
 792                return -1;
 793
 794        /* find the end of the word */
 795        *end = *begin + 1;
 796        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 797                (*end)++;
 798
 799        return 0;
 800}
 801
 802/*
 803 * This function splits the words in buffer->text, stores the list with
 804 * newline separator into out, and saves the offsets of the original words
 805 * in buffer->orig.
 806 */
 807static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 808                regex_t *word_regex)
 809{
 810        int i, j;
 811        long alloc = 0;
 812
 813        out->size = 0;
 814        out->ptr = NULL;
 815
 816        /* fake an empty "0th" word */
 817        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 818        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 819        buffer->orig_nr = 1;
 820
 821        for (i = 0; i < buffer->text.size; i++) {
 822                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 823                        return;
 824
 825                /* store original boundaries */
 826                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 827                                buffer->orig_alloc);
 828                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 829                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 830                buffer->orig_nr++;
 831
 832                /* store one word */
 833                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 834                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 835                out->ptr[out->size + j - i] = '\n';
 836                out->size += j - i + 1;
 837
 838                i = j - 1;
 839        }
 840}
 841
 842/* this executes the word diff on the accumulated buffers */
 843static void diff_words_show(struct diff_words_data *diff_words)
 844{
 845        xpparam_t xpp;
 846        xdemitconf_t xecfg;
 847        mmfile_t minus, plus;
 848        struct diff_words_style *style = diff_words->style;
 849
 850        struct diff_options *opt = diff_words->opt;
 851        struct strbuf *msgbuf;
 852        char *line_prefix = "";
 853
 854        assert(opt);
 855        if (opt->output_prefix) {
 856                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 857                line_prefix = msgbuf->buf;
 858        }
 859
 860        /* special case: only removal */
 861        if (!diff_words->plus.text.size) {
 862                fputs(line_prefix, diff_words->opt->file);
 863                fn_out_diff_words_write_helper(diff_words->opt->file,
 864                        &style->old, style->newline,
 865                        diff_words->minus.text.size,
 866                        diff_words->minus.text.ptr, line_prefix);
 867                diff_words->minus.text.size = 0;
 868                return;
 869        }
 870
 871        diff_words->current_plus = diff_words->plus.text.ptr;
 872        diff_words->last_minus = 0;
 873
 874        memset(&xpp, 0, sizeof(xpp));
 875        memset(&xecfg, 0, sizeof(xecfg));
 876        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 877        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 878        xpp.flags = XDF_NEED_MINIMAL;
 879        /* as only the hunk header will be parsed, we need a 0-context */
 880        xecfg.ctxlen = 0;
 881        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 882                      &xpp, &xecfg);
 883        free(minus.ptr);
 884        free(plus.ptr);
 885        if (diff_words->current_plus != diff_words->plus.text.ptr +
 886                        diff_words->plus.text.size) {
 887                if (color_words_output_graph_prefix(diff_words))
 888                        fputs(line_prefix, diff_words->opt->file);
 889                fn_out_diff_words_write_helper(diff_words->opt->file,
 890                        &style->ctx, style->newline,
 891                        diff_words->plus.text.ptr + diff_words->plus.text.size
 892                        - diff_words->current_plus, diff_words->current_plus,
 893                        line_prefix);
 894        }
 895        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 896}
 897
 898/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 899static void diff_words_flush(struct emit_callback *ecbdata)
 900{
 901        if (ecbdata->diff_words->minus.text.size ||
 902            ecbdata->diff_words->plus.text.size)
 903                diff_words_show(ecbdata->diff_words);
 904}
 905
 906static void free_diff_words_data(struct emit_callback *ecbdata)
 907{
 908        if (ecbdata->diff_words) {
 909                diff_words_flush(ecbdata);
 910                free (ecbdata->diff_words->minus.text.ptr);
 911                free (ecbdata->diff_words->minus.orig);
 912                free (ecbdata->diff_words->plus.text.ptr);
 913                free (ecbdata->diff_words->plus.orig);
 914                free(ecbdata->diff_words->word_regex);
 915                free(ecbdata->diff_words);
 916                ecbdata->diff_words = NULL;
 917        }
 918}
 919
 920const char *diff_get_color(int diff_use_color, enum color_diff ix)
 921{
 922        if (diff_use_color)
 923                return diff_colors[ix];
 924        return "";
 925}
 926
 927static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
 928{
 929        const char *cp;
 930        unsigned long allot;
 931        size_t l = len;
 932
 933        if (ecb->truncate)
 934                return ecb->truncate(line, len);
 935        cp = line;
 936        allot = l;
 937        while (0 < l) {
 938                (void) utf8_width(&cp, &l);
 939                if (!cp)
 940                        break; /* truncated in the middle? */
 941        }
 942        return allot - l;
 943}
 944
 945static void find_lno(const char *line, struct emit_callback *ecbdata)
 946{
 947        const char *p;
 948        ecbdata->lno_in_preimage = 0;
 949        ecbdata->lno_in_postimage = 0;
 950        p = strchr(line, '-');
 951        if (!p)
 952                return; /* cannot happen */
 953        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
 954        p = strchr(p, '+');
 955        if (!p)
 956                return; /* cannot happen */
 957        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
 958}
 959
 960static void fn_out_consume(void *priv, char *line, unsigned long len)
 961{
 962        struct emit_callback *ecbdata = priv;
 963        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 964        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 965        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 966        struct diff_options *o = ecbdata->opt;
 967        char *line_prefix = "";
 968        struct strbuf *msgbuf;
 969
 970        if (o && o->output_prefix) {
 971                msgbuf = o->output_prefix(o, o->output_prefix_data);
 972                line_prefix = msgbuf->buf;
 973        }
 974
 975        if (ecbdata->header) {
 976                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
 977                strbuf_reset(ecbdata->header);
 978                ecbdata->header = NULL;
 979        }
 980        *(ecbdata->found_changesp) = 1;
 981
 982        if (ecbdata->label_path[0]) {
 983                const char *name_a_tab, *name_b_tab;
 984
 985                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 986                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 987
 988                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
 989                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
 990                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
 991                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
 992                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 993        }
 994
 995        if (diff_suppress_blank_empty
 996            && len == 2 && line[0] == ' ' && line[1] == '\n') {
 997                line[0] = '\n';
 998                len = 1;
 999        }
1000
1001        if (line[0] == '@') {
1002                if (ecbdata->diff_words)
1003                        diff_words_flush(ecbdata);
1004                len = sane_truncate_line(ecbdata, line, len);
1005                find_lno(line, ecbdata);
1006                emit_hunk_header(ecbdata, line, len);
1007                if (line[len-1] != '\n')
1008                        putc('\n', ecbdata->opt->file);
1009                return;
1010        }
1011
1012        if (len < 1) {
1013                emit_line(ecbdata->opt, reset, reset, line, len);
1014                if (ecbdata->diff_words
1015                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1016                        fputs("~\n", ecbdata->opt->file);
1017                return;
1018        }
1019
1020        if (ecbdata->diff_words) {
1021                if (line[0] == '-') {
1022                        diff_words_append(line, len,
1023                                          &ecbdata->diff_words->minus);
1024                        return;
1025                } else if (line[0] == '+') {
1026                        diff_words_append(line, len,
1027                                          &ecbdata->diff_words->plus);
1028                        return;
1029                }
1030                diff_words_flush(ecbdata);
1031                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1032                        emit_line(ecbdata->opt, plain, reset, line, len);
1033                        fputs("~\n", ecbdata->opt->file);
1034                } else {
1035                        /* don't print the prefix character */
1036                        emit_line(ecbdata->opt, plain, reset, line+1, len-1);
1037                }
1038                return;
1039        }
1040
1041        if (line[0] != '+') {
1042                const char *color =
1043                        diff_get_color(ecbdata->color_diff,
1044                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1045                ecbdata->lno_in_preimage++;
1046                if (line[0] == ' ')
1047                        ecbdata->lno_in_postimage++;
1048                emit_line(ecbdata->opt, color, reset, line, len);
1049        } else {
1050                ecbdata->lno_in_postimage++;
1051                emit_add_line(reset, ecbdata, line + 1, len - 1);
1052        }
1053}
1054
1055static char *pprint_rename(const char *a, const char *b)
1056{
1057        const char *old = a;
1058        const char *new = b;
1059        struct strbuf name = STRBUF_INIT;
1060        int pfx_length, sfx_length;
1061        int len_a = strlen(a);
1062        int len_b = strlen(b);
1063        int a_midlen, b_midlen;
1064        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1065        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1066
1067        if (qlen_a || qlen_b) {
1068                quote_c_style(a, &name, NULL, 0);
1069                strbuf_addstr(&name, " => ");
1070                quote_c_style(b, &name, NULL, 0);
1071                return strbuf_detach(&name, NULL);
1072        }
1073
1074        /* Find common prefix */
1075        pfx_length = 0;
1076        while (*old && *new && *old == *new) {
1077                if (*old == '/')
1078                        pfx_length = old - a + 1;
1079                old++;
1080                new++;
1081        }
1082
1083        /* Find common suffix */
1084        old = a + len_a;
1085        new = b + len_b;
1086        sfx_length = 0;
1087        while (a <= old && b <= new && *old == *new) {
1088                if (*old == '/')
1089                        sfx_length = len_a - (old - a);
1090                old--;
1091                new--;
1092        }
1093
1094        /*
1095         * pfx{mid-a => mid-b}sfx
1096         * {pfx-a => pfx-b}sfx
1097         * pfx{sfx-a => sfx-b}
1098         * name-a => name-b
1099         */
1100        a_midlen = len_a - pfx_length - sfx_length;
1101        b_midlen = len_b - pfx_length - sfx_length;
1102        if (a_midlen < 0)
1103                a_midlen = 0;
1104        if (b_midlen < 0)
1105                b_midlen = 0;
1106
1107        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1108        if (pfx_length + sfx_length) {
1109                strbuf_add(&name, a, pfx_length);
1110                strbuf_addch(&name, '{');
1111        }
1112        strbuf_add(&name, a + pfx_length, a_midlen);
1113        strbuf_addstr(&name, " => ");
1114        strbuf_add(&name, b + pfx_length, b_midlen);
1115        if (pfx_length + sfx_length) {
1116                strbuf_addch(&name, '}');
1117                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1118        }
1119        return strbuf_detach(&name, NULL);
1120}
1121
1122struct diffstat_t {
1123        int nr;
1124        int alloc;
1125        struct diffstat_file {
1126                char *from_name;
1127                char *name;
1128                char *print_name;
1129                unsigned is_unmerged:1;
1130                unsigned is_binary:1;
1131                unsigned is_renamed:1;
1132                uintmax_t added, deleted;
1133        } **files;
1134};
1135
1136static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1137                                          const char *name_a,
1138                                          const char *name_b)
1139{
1140        struct diffstat_file *x;
1141        x = xcalloc(sizeof (*x), 1);
1142        if (diffstat->nr == diffstat->alloc) {
1143                diffstat->alloc = alloc_nr(diffstat->alloc);
1144                diffstat->files = xrealloc(diffstat->files,
1145                                diffstat->alloc * sizeof(x));
1146        }
1147        diffstat->files[diffstat->nr++] = x;
1148        if (name_b) {
1149                x->from_name = xstrdup(name_a);
1150                x->name = xstrdup(name_b);
1151                x->is_renamed = 1;
1152        }
1153        else {
1154                x->from_name = NULL;
1155                x->name = xstrdup(name_a);
1156        }
1157        return x;
1158}
1159
1160static void diffstat_consume(void *priv, char *line, unsigned long len)
1161{
1162        struct diffstat_t *diffstat = priv;
1163        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1164
1165        if (line[0] == '+')
1166                x->added++;
1167        else if (line[0] == '-')
1168                x->deleted++;
1169}
1170
1171const char mime_boundary_leader[] = "------------";
1172
1173static int scale_linear(int it, int width, int max_change)
1174{
1175        /*
1176         * make sure that at least one '-' is printed if there were deletions,
1177         * and likewise for '+'.
1178         */
1179        if (max_change < 2)
1180                return it;
1181        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
1182}
1183
1184static void show_name(FILE *file,
1185                      const char *prefix, const char *name, int len)
1186{
1187        fprintf(file, " %s%-*s |", prefix, len, name);
1188}
1189
1190static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1191{
1192        if (cnt <= 0)
1193                return;
1194        fprintf(file, "%s", set);
1195        while (cnt--)
1196                putc(ch, file);
1197        fprintf(file, "%s", reset);
1198}
1199
1200static void fill_print_name(struct diffstat_file *file)
1201{
1202        char *pname;
1203
1204        if (file->print_name)
1205                return;
1206
1207        if (!file->is_renamed) {
1208                struct strbuf buf = STRBUF_INIT;
1209                if (quote_c_style(file->name, &buf, NULL, 0)) {
1210                        pname = strbuf_detach(&buf, NULL);
1211                } else {
1212                        pname = file->name;
1213                        strbuf_release(&buf);
1214                }
1215        } else {
1216                pname = pprint_rename(file->from_name, file->name);
1217        }
1218        file->print_name = pname;
1219}
1220
1221static void show_stats(struct diffstat_t *data, struct diff_options *options)
1222{
1223        int i, len, add, del, adds = 0, dels = 0;
1224        uintmax_t max_change = 0, max_len = 0;
1225        int total_files = data->nr;
1226        int width, name_width;
1227        const char *reset, *set, *add_c, *del_c;
1228        const char *line_prefix = "";
1229        struct strbuf *msg = NULL;
1230
1231        if (data->nr == 0)
1232                return;
1233
1234        if (options->output_prefix) {
1235                msg = options->output_prefix(options, options->output_prefix_data);
1236                line_prefix = msg->buf;
1237        }
1238
1239        width = options->stat_width ? options->stat_width : 80;
1240        name_width = options->stat_name_width ? options->stat_name_width : 50;
1241
1242        /* Sanity: give at least 5 columns to the graph,
1243         * but leave at least 10 columns for the name.
1244         */
1245        if (width < 25)
1246                width = 25;
1247        if (name_width < 10)
1248                name_width = 10;
1249        else if (width < name_width + 15)
1250                name_width = width - 15;
1251
1252        /* Find the longest filename and max number of changes */
1253        reset = diff_get_color_opt(options, DIFF_RESET);
1254        set   = diff_get_color_opt(options, DIFF_PLAIN);
1255        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1256        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1257
1258        for (i = 0; i < data->nr; i++) {
1259                struct diffstat_file *file = data->files[i];
1260                uintmax_t change = file->added + file->deleted;
1261                fill_print_name(file);
1262                len = strlen(file->print_name);
1263                if (max_len < len)
1264                        max_len = len;
1265
1266                if (file->is_binary || file->is_unmerged)
1267                        continue;
1268                if (max_change < change)
1269                        max_change = change;
1270        }
1271
1272        /* Compute the width of the graph part;
1273         * 10 is for one blank at the beginning of the line plus
1274         * " | count " between the name and the graph.
1275         *
1276         * From here on, name_width is the width of the name area,
1277         * and width is the width of the graph area.
1278         */
1279        name_width = (name_width < max_len) ? name_width : max_len;
1280        if (width < (name_width + 10) + max_change)
1281                width = width - (name_width + 10);
1282        else
1283                width = max_change;
1284
1285        for (i = 0; i < data->nr; i++) {
1286                const char *prefix = "";
1287                char *name = data->files[i]->print_name;
1288                uintmax_t added = data->files[i]->added;
1289                uintmax_t deleted = data->files[i]->deleted;
1290                int name_len;
1291
1292                /*
1293                 * "scale" the filename
1294                 */
1295                len = name_width;
1296                name_len = strlen(name);
1297                if (name_width < name_len) {
1298                        char *slash;
1299                        prefix = "...";
1300                        len -= 3;
1301                        name += name_len - len;
1302                        slash = strchr(name, '/');
1303                        if (slash)
1304                                name = slash;
1305                }
1306
1307                if (data->files[i]->is_binary) {
1308                        fprintf(options->file, "%s", line_prefix);
1309                        show_name(options->file, prefix, name, len);
1310                        fprintf(options->file, "  Bin ");
1311                        fprintf(options->file, "%s%"PRIuMAX"%s",
1312                                del_c, deleted, reset);
1313                        fprintf(options->file, " -> ");
1314                        fprintf(options->file, "%s%"PRIuMAX"%s",
1315                                add_c, added, reset);
1316                        fprintf(options->file, " bytes");
1317                        fprintf(options->file, "\n");
1318                        continue;
1319                }
1320                else if (data->files[i]->is_unmerged) {
1321                        fprintf(options->file, "%s", line_prefix);
1322                        show_name(options->file, prefix, name, len);
1323                        fprintf(options->file, "  Unmerged\n");
1324                        continue;
1325                }
1326                else if (!data->files[i]->is_renamed &&
1327                         (added + deleted == 0)) {
1328                        total_files--;
1329                        continue;
1330                }
1331
1332                /*
1333                 * scale the add/delete
1334                 */
1335                add = added;
1336                del = deleted;
1337                adds += add;
1338                dels += del;
1339
1340                if (width <= max_change) {
1341                        add = scale_linear(add, width, max_change);
1342                        del = scale_linear(del, width, max_change);
1343                }
1344                fprintf(options->file, "%s", line_prefix);
1345                show_name(options->file, prefix, name, len);
1346                fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1347                                added + deleted ? " " : "");
1348                show_graph(options->file, '+', add, add_c, reset);
1349                show_graph(options->file, '-', del, del_c, reset);
1350                fprintf(options->file, "\n");
1351        }
1352        fprintf(options->file, "%s", line_prefix);
1353        fprintf(options->file,
1354               " %d files changed, %d insertions(+), %d deletions(-)\n",
1355               total_files, adds, dels);
1356}
1357
1358static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1359{
1360        int i, adds = 0, dels = 0, total_files = data->nr;
1361
1362        if (data->nr == 0)
1363                return;
1364
1365        for (i = 0; i < data->nr; i++) {
1366                if (!data->files[i]->is_binary &&
1367                    !data->files[i]->is_unmerged) {
1368                        int added = data->files[i]->added;
1369                        int deleted= data->files[i]->deleted;
1370                        if (!data->files[i]->is_renamed &&
1371                            (added + deleted == 0)) {
1372                                total_files--;
1373                        } else {
1374                                adds += added;
1375                                dels += deleted;
1376                        }
1377                }
1378        }
1379        if (options->output_prefix) {
1380                struct strbuf *msg = NULL;
1381                msg = options->output_prefix(options,
1382                                options->output_prefix_data);
1383                fprintf(options->file, "%s", msg->buf);
1384        }
1385        fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1386               total_files, adds, dels);
1387}
1388
1389static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1390{
1391        int i;
1392
1393        if (data->nr == 0)
1394                return;
1395
1396        for (i = 0; i < data->nr; i++) {
1397                struct diffstat_file *file = data->files[i];
1398
1399                if (options->output_prefix) {
1400                        struct strbuf *msg = NULL;
1401                        msg = options->output_prefix(options,
1402                                        options->output_prefix_data);
1403                        fprintf(options->file, "%s", msg->buf);
1404                }
1405
1406                if (file->is_binary)
1407                        fprintf(options->file, "-\t-\t");
1408                else
1409                        fprintf(options->file,
1410                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1411                                file->added, file->deleted);
1412                if (options->line_termination) {
1413                        fill_print_name(file);
1414                        if (!file->is_renamed)
1415                                write_name_quoted(file->name, options->file,
1416                                                  options->line_termination);
1417                        else {
1418                                fputs(file->print_name, options->file);
1419                                putc(options->line_termination, options->file);
1420                        }
1421                } else {
1422                        if (file->is_renamed) {
1423                                putc('\0', options->file);
1424                                write_name_quoted(file->from_name, options->file, '\0');
1425                        }
1426                        write_name_quoted(file->name, options->file, '\0');
1427                }
1428        }
1429}
1430
1431struct dirstat_file {
1432        const char *name;
1433        unsigned long changed;
1434};
1435
1436struct dirstat_dir {
1437        struct dirstat_file *files;
1438        int alloc, nr, percent, cumulative;
1439};
1440
1441static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1442                unsigned long changed, const char *base, int baselen)
1443{
1444        unsigned long this_dir = 0;
1445        unsigned int sources = 0;
1446        const char *line_prefix = "";
1447        struct strbuf *msg = NULL;
1448
1449        if (opt->output_prefix) {
1450                msg = opt->output_prefix(opt, opt->output_prefix_data);
1451                line_prefix = msg->buf;
1452        }
1453
1454        while (dir->nr) {
1455                struct dirstat_file *f = dir->files;
1456                int namelen = strlen(f->name);
1457                unsigned long this;
1458                char *slash;
1459
1460                if (namelen < baselen)
1461                        break;
1462                if (memcmp(f->name, base, baselen))
1463                        break;
1464                slash = strchr(f->name + baselen, '/');
1465                if (slash) {
1466                        int newbaselen = slash + 1 - f->name;
1467                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1468                        sources++;
1469                } else {
1470                        this = f->changed;
1471                        dir->files++;
1472                        dir->nr--;
1473                        sources += 2;
1474                }
1475                this_dir += this;
1476        }
1477
1478        /*
1479         * We don't report dirstat's for
1480         *  - the top level
1481         *  - or cases where everything came from a single directory
1482         *    under this directory (sources == 1).
1483         */
1484        if (baselen && sources != 1) {
1485                int permille = this_dir * 1000 / changed;
1486                if (permille) {
1487                        int percent = permille / 10;
1488                        if (percent >= dir->percent) {
1489                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1490                                        percent, permille % 10, baselen, base);
1491                                if (!dir->cumulative)
1492                                        return 0;
1493                        }
1494                }
1495        }
1496        return this_dir;
1497}
1498
1499static int dirstat_compare(const void *_a, const void *_b)
1500{
1501        const struct dirstat_file *a = _a;
1502        const struct dirstat_file *b = _b;
1503        return strcmp(a->name, b->name);
1504}
1505
1506static void show_dirstat(struct diff_options *options)
1507{
1508        int i;
1509        unsigned long changed;
1510        struct dirstat_dir dir;
1511        struct diff_queue_struct *q = &diff_queued_diff;
1512
1513        dir.files = NULL;
1514        dir.alloc = 0;
1515        dir.nr = 0;
1516        dir.percent = options->dirstat_percent;
1517        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1518
1519        changed = 0;
1520        for (i = 0; i < q->nr; i++) {
1521                struct diff_filepair *p = q->queue[i];
1522                const char *name;
1523                unsigned long copied, added, damage;
1524
1525                name = p->one->path ? p->one->path : p->two->path;
1526
1527                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1528                        diff_populate_filespec(p->one, 0);
1529                        diff_populate_filespec(p->two, 0);
1530                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1531                                               &copied, &added);
1532                        diff_free_filespec_data(p->one);
1533                        diff_free_filespec_data(p->two);
1534                } else if (DIFF_FILE_VALID(p->one)) {
1535                        diff_populate_filespec(p->one, 1);
1536                        copied = added = 0;
1537                        diff_free_filespec_data(p->one);
1538                } else if (DIFF_FILE_VALID(p->two)) {
1539                        diff_populate_filespec(p->two, 1);
1540                        copied = 0;
1541                        added = p->two->size;
1542                        diff_free_filespec_data(p->two);
1543                } else
1544                        continue;
1545
1546                /*
1547                 * Original minus copied is the removed material,
1548                 * added is the new material.  They are both damages
1549                 * made to the preimage. In --dirstat-by-file mode, count
1550                 * damaged files, not damaged lines. This is done by
1551                 * counting only a single damaged line per file.
1552                 */
1553                damage = (p->one->size - copied) + added;
1554                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE) && damage > 0)
1555                        damage = 1;
1556
1557                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1558                dir.files[dir.nr].name = name;
1559                dir.files[dir.nr].changed = damage;
1560                changed += damage;
1561                dir.nr++;
1562        }
1563
1564        /* This can happen even with many files, if everything was renames */
1565        if (!changed)
1566                return;
1567
1568        /* Show all directories with more than x% of the changes */
1569        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1570        gather_dirstat(options, &dir, changed, "", 0);
1571}
1572
1573static void free_diffstat_info(struct diffstat_t *diffstat)
1574{
1575        int i;
1576        for (i = 0; i < diffstat->nr; i++) {
1577                struct diffstat_file *f = diffstat->files[i];
1578                if (f->name != f->print_name)
1579                        free(f->print_name);
1580                free(f->name);
1581                free(f->from_name);
1582                free(f);
1583        }
1584        free(diffstat->files);
1585}
1586
1587struct checkdiff_t {
1588        const char *filename;
1589        int lineno;
1590        int conflict_marker_size;
1591        struct diff_options *o;
1592        unsigned ws_rule;
1593        unsigned status;
1594};
1595
1596static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1597{
1598        char firstchar;
1599        int cnt;
1600
1601        if (len < marker_size + 1)
1602                return 0;
1603        firstchar = line[0];
1604        switch (firstchar) {
1605        case '=': case '>': case '<': case '|':
1606                break;
1607        default:
1608                return 0;
1609        }
1610        for (cnt = 1; cnt < marker_size; cnt++)
1611                if (line[cnt] != firstchar)
1612                        return 0;
1613        /* line[1] thru line[marker_size-1] are same as firstchar */
1614        if (len < marker_size + 1 || !isspace(line[marker_size]))
1615                return 0;
1616        return 1;
1617}
1618
1619static void checkdiff_consume(void *priv, char *line, unsigned long len)
1620{
1621        struct checkdiff_t *data = priv;
1622        int color_diff = DIFF_OPT_TST(data->o, COLOR_DIFF);
1623        int marker_size = data->conflict_marker_size;
1624        const char *ws = diff_get_color(color_diff, DIFF_WHITESPACE);
1625        const char *reset = diff_get_color(color_diff, DIFF_RESET);
1626        const char *set = diff_get_color(color_diff, DIFF_FILE_NEW);
1627        char *err;
1628        char *line_prefix = "";
1629        struct strbuf *msgbuf;
1630
1631        assert(data->o);
1632        if (data->o->output_prefix) {
1633                msgbuf = data->o->output_prefix(data->o,
1634                        data->o->output_prefix_data);
1635                line_prefix = msgbuf->buf;
1636        }
1637
1638        if (line[0] == '+') {
1639                unsigned bad;
1640                data->lineno++;
1641                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1642                        data->status |= 1;
1643                        fprintf(data->o->file,
1644                                "%s%s:%d: leftover conflict marker\n",
1645                                line_prefix, data->filename, data->lineno);
1646                }
1647                bad = ws_check(line + 1, len - 1, data->ws_rule);
1648                if (!bad)
1649                        return;
1650                data->status |= bad;
1651                err = whitespace_error_string(bad);
1652                fprintf(data->o->file, "%s%s:%d: %s.\n",
1653                        line_prefix, data->filename, data->lineno, err);
1654                free(err);
1655                emit_line(data->o, set, reset, line, 1);
1656                ws_check_emit(line + 1, len - 1, data->ws_rule,
1657                              data->o->file, set, reset, ws);
1658        } else if (line[0] == ' ') {
1659                data->lineno++;
1660        } else if (line[0] == '@') {
1661                char *plus = strchr(line, '+');
1662                if (plus)
1663                        data->lineno = strtol(plus, NULL, 10) - 1;
1664                else
1665                        die("invalid diff");
1666        }
1667}
1668
1669static unsigned char *deflate_it(char *data,
1670                                 unsigned long size,
1671                                 unsigned long *result_size)
1672{
1673        int bound;
1674        unsigned char *deflated;
1675        z_stream stream;
1676
1677        memset(&stream, 0, sizeof(stream));
1678        deflateInit(&stream, zlib_compression_level);
1679        bound = deflateBound(&stream, size);
1680        deflated = xmalloc(bound);
1681        stream.next_out = deflated;
1682        stream.avail_out = bound;
1683
1684        stream.next_in = (unsigned char *)data;
1685        stream.avail_in = size;
1686        while (deflate(&stream, Z_FINISH) == Z_OK)
1687                ; /* nothing */
1688        deflateEnd(&stream);
1689        *result_size = stream.total_out;
1690        return deflated;
1691}
1692
1693static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1694{
1695        void *cp;
1696        void *delta;
1697        void *deflated;
1698        void *data;
1699        unsigned long orig_size;
1700        unsigned long delta_size;
1701        unsigned long deflate_size;
1702        unsigned long data_size;
1703
1704        /* We could do deflated delta, or we could do just deflated two,
1705         * whichever is smaller.
1706         */
1707        delta = NULL;
1708        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1709        if (one->size && two->size) {
1710                delta = diff_delta(one->ptr, one->size,
1711                                   two->ptr, two->size,
1712                                   &delta_size, deflate_size);
1713                if (delta) {
1714                        void *to_free = delta;
1715                        orig_size = delta_size;
1716                        delta = deflate_it(delta, delta_size, &delta_size);
1717                        free(to_free);
1718                }
1719        }
1720
1721        if (delta && delta_size < deflate_size) {
1722                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
1723                free(deflated);
1724                data = delta;
1725                data_size = delta_size;
1726        }
1727        else {
1728                fprintf(file, "%sliteral %lu\n", prefix, two->size);
1729                free(delta);
1730                data = deflated;
1731                data_size = deflate_size;
1732        }
1733
1734        /* emit data encoded in base85 */
1735        cp = data;
1736        while (data_size) {
1737                int bytes = (52 < data_size) ? 52 : data_size;
1738                char line[70];
1739                data_size -= bytes;
1740                if (bytes <= 26)
1741                        line[0] = bytes + 'A' - 1;
1742                else
1743                        line[0] = bytes - 26 + 'a' - 1;
1744                encode_85(line + 1, cp, bytes);
1745                cp = (char *) cp + bytes;
1746                fprintf(file, "%s", prefix);
1747                fputs(line, file);
1748                fputc('\n', file);
1749        }
1750        fprintf(file, "%s\n", prefix);
1751        free(data);
1752}
1753
1754static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1755{
1756        fprintf(file, "%sGIT binary patch\n", prefix);
1757        emit_binary_diff_body(file, one, two, prefix);
1758        emit_binary_diff_body(file, two, one, prefix);
1759}
1760
1761static void diff_filespec_load_driver(struct diff_filespec *one)
1762{
1763        if (!one->driver)
1764                one->driver = userdiff_find_by_path(one->path);
1765        if (!one->driver)
1766                one->driver = userdiff_find_by_name("default");
1767}
1768
1769int diff_filespec_is_binary(struct diff_filespec *one)
1770{
1771        if (one->is_binary == -1) {
1772                diff_filespec_load_driver(one);
1773                if (one->driver->binary != -1)
1774                        one->is_binary = one->driver->binary;
1775                else {
1776                        if (!one->data && DIFF_FILE_VALID(one))
1777                                diff_populate_filespec(one, 0);
1778                        if (one->data)
1779                                one->is_binary = buffer_is_binary(one->data,
1780                                                one->size);
1781                        if (one->is_binary == -1)
1782                                one->is_binary = 0;
1783                }
1784        }
1785        return one->is_binary;
1786}
1787
1788static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
1789{
1790        diff_filespec_load_driver(one);
1791        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
1792}
1793
1794static const char *userdiff_word_regex(struct diff_filespec *one)
1795{
1796        diff_filespec_load_driver(one);
1797        return one->driver->word_regex;
1798}
1799
1800void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
1801{
1802        if (!options->a_prefix)
1803                options->a_prefix = a;
1804        if (!options->b_prefix)
1805                options->b_prefix = b;
1806}
1807
1808static struct userdiff_driver *get_textconv(struct diff_filespec *one)
1809{
1810        if (!DIFF_FILE_VALID(one))
1811                return NULL;
1812        if (!S_ISREG(one->mode))
1813                return NULL;
1814        diff_filespec_load_driver(one);
1815        if (!one->driver->textconv)
1816                return NULL;
1817
1818        if (one->driver->textconv_want_cache && !one->driver->textconv_cache) {
1819                struct notes_cache *c = xmalloc(sizeof(*c));
1820                struct strbuf name = STRBUF_INIT;
1821
1822                strbuf_addf(&name, "textconv/%s", one->driver->name);
1823                notes_cache_init(c, name.buf, one->driver->textconv);
1824                one->driver->textconv_cache = c;
1825        }
1826
1827        return one->driver;
1828}
1829
1830static void builtin_diff(const char *name_a,
1831                         const char *name_b,
1832                         struct diff_filespec *one,
1833                         struct diff_filespec *two,
1834                         const char *xfrm_msg,
1835                         struct diff_options *o,
1836                         int complete_rewrite)
1837{
1838        mmfile_t mf1, mf2;
1839        const char *lbl[2];
1840        char *a_one, *b_two;
1841        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1842        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1843        const char *a_prefix, *b_prefix;
1844        struct userdiff_driver *textconv_one = NULL;
1845        struct userdiff_driver *textconv_two = NULL;
1846        struct strbuf header = STRBUF_INIT;
1847        struct strbuf *msgbuf;
1848        char *line_prefix = "";
1849
1850        if (o->output_prefix) {
1851                msgbuf = o->output_prefix(o, o->output_prefix_data);
1852                line_prefix = msgbuf->buf;
1853        }
1854
1855        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
1856                        (!one->mode || S_ISGITLINK(one->mode)) &&
1857                        (!two->mode || S_ISGITLINK(two->mode))) {
1858                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
1859                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
1860                show_submodule_summary(o->file, one ? one->path : two->path,
1861                                one->sha1, two->sha1, two->dirty_submodule,
1862                                del, add, reset);
1863                return;
1864        }
1865
1866        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
1867                textconv_one = get_textconv(one);
1868                textconv_two = get_textconv(two);
1869        }
1870
1871        diff_set_mnemonic_prefix(o, "a/", "b/");
1872        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
1873                a_prefix = o->b_prefix;
1874                b_prefix = o->a_prefix;
1875        } else {
1876                a_prefix = o->a_prefix;
1877                b_prefix = o->b_prefix;
1878        }
1879
1880        /* Never use a non-valid filename anywhere if at all possible */
1881        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
1882        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
1883
1884        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
1885        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
1886        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1887        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1888        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
1889        if (lbl[0][0] == '/') {
1890                /* /dev/null */
1891                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
1892                if (xfrm_msg && xfrm_msg[0])
1893                        strbuf_addf(&header, "%s%s%s\n", set, xfrm_msg, reset);
1894        }
1895        else if (lbl[1][0] == '/') {
1896                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
1897                if (xfrm_msg && xfrm_msg[0])
1898                        strbuf_addf(&header, "%s%s%s\n", set, xfrm_msg, reset);
1899        }
1900        else {
1901                if (one->mode != two->mode) {
1902                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
1903                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
1904                }
1905                if (xfrm_msg && xfrm_msg[0])
1906                        strbuf_addf(&header, "%s%s%s\n", set, xfrm_msg, reset);
1907
1908                /*
1909                 * we do not run diff between different kind
1910                 * of objects.
1911                 */
1912                if ((one->mode ^ two->mode) & S_IFMT)
1913                        goto free_ab_and_return;
1914                if (complete_rewrite &&
1915                    (textconv_one || !diff_filespec_is_binary(one)) &&
1916                    (textconv_two || !diff_filespec_is_binary(two))) {
1917                        fprintf(o->file, "%s", header.buf);
1918                        strbuf_reset(&header);
1919                        emit_rewrite_diff(name_a, name_b, one, two,
1920                                                textconv_one, textconv_two, o);
1921                        o->found_changes = 1;
1922                        goto free_ab_and_return;
1923                }
1924        }
1925
1926        if (!DIFF_OPT_TST(o, TEXT) &&
1927            ( (!textconv_one && diff_filespec_is_binary(one)) ||
1928              (!textconv_two && diff_filespec_is_binary(two)) )) {
1929                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1930                        die("unable to read files to diff");
1931                /* Quite common confusing case */
1932                if (mf1.size == mf2.size &&
1933                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1934                        goto free_ab_and_return;
1935                fprintf(o->file, "%s", header.buf);
1936                strbuf_reset(&header);
1937                if (DIFF_OPT_TST(o, BINARY))
1938                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
1939                else
1940                        fprintf(o->file, "%sBinary files %s and %s differ\n",
1941                                line_prefix, lbl[0], lbl[1]);
1942                o->found_changes = 1;
1943        }
1944        else {
1945                /* Crazy xdl interfaces.. */
1946                const char *diffopts = getenv("GIT_DIFF_OPTS");
1947                xpparam_t xpp;
1948                xdemitconf_t xecfg;
1949                struct emit_callback ecbdata;
1950                const struct userdiff_funcname *pe;
1951
1952                if (!DIFF_XDL_TST(o, WHITESPACE_FLAGS)) {
1953                        fprintf(o->file, "%s", header.buf);
1954                        strbuf_reset(&header);
1955                }
1956
1957                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
1958                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
1959
1960                pe = diff_funcname_pattern(one);
1961                if (!pe)
1962                        pe = diff_funcname_pattern(two);
1963
1964                memset(&xpp, 0, sizeof(xpp));
1965                memset(&xecfg, 0, sizeof(xecfg));
1966                memset(&ecbdata, 0, sizeof(ecbdata));
1967                ecbdata.label_path = lbl;
1968                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1969                ecbdata.found_changesp = &o->found_changes;
1970                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1971                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
1972                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
1973                ecbdata.opt = o;
1974                ecbdata.header = header.len ? &header : NULL;
1975                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1976                xecfg.ctxlen = o->context;
1977                xecfg.interhunkctxlen = o->interhunkcontext;
1978                xecfg.flags = XDL_EMIT_FUNCNAMES;
1979                if (pe)
1980                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
1981                if (!diffopts)
1982                        ;
1983                else if (!prefixcmp(diffopts, "--unified="))
1984                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1985                else if (!prefixcmp(diffopts, "-u"))
1986                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1987                if (o->word_diff) {
1988                        int i;
1989
1990                        ecbdata.diff_words =
1991                                xcalloc(1, sizeof(struct diff_words_data));
1992                        ecbdata.diff_words->type = o->word_diff;
1993                        ecbdata.diff_words->opt = o;
1994                        if (!o->word_regex)
1995                                o->word_regex = userdiff_word_regex(one);
1996                        if (!o->word_regex)
1997                                o->word_regex = userdiff_word_regex(two);
1998                        if (!o->word_regex)
1999                                o->word_regex = diff_word_regex_cfg;
2000                        if (o->word_regex) {
2001                                ecbdata.diff_words->word_regex = (regex_t *)
2002                                        xmalloc(sizeof(regex_t));
2003                                if (regcomp(ecbdata.diff_words->word_regex,
2004                                                o->word_regex,
2005                                                REG_EXTENDED | REG_NEWLINE))
2006                                        die ("Invalid regular expression: %s",
2007                                                        o->word_regex);
2008                        }
2009                        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2010                                if (o->word_diff == diff_words_styles[i].type) {
2011                                        ecbdata.diff_words->style =
2012                                                &diff_words_styles[i];
2013                                        break;
2014                                }
2015                        }
2016                        if (DIFF_OPT_TST(o, COLOR_DIFF)) {
2017                                struct diff_words_style *st = ecbdata.diff_words->style;
2018                                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2019                                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2020                                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
2021                        }
2022                }
2023                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2024                              &xpp, &xecfg);
2025                if (o->word_diff)
2026                        free_diff_words_data(&ecbdata);
2027                if (textconv_one)
2028                        free(mf1.ptr);
2029                if (textconv_two)
2030                        free(mf2.ptr);
2031                xdiff_clear_find_func(&xecfg);
2032        }
2033
2034 free_ab_and_return:
2035        strbuf_release(&header);
2036        diff_free_filespec_data(one);
2037        diff_free_filespec_data(two);
2038        free(a_one);
2039        free(b_two);
2040        return;
2041}
2042
2043static void builtin_diffstat(const char *name_a, const char *name_b,
2044                             struct diff_filespec *one,
2045                             struct diff_filespec *two,
2046                             struct diffstat_t *diffstat,
2047                             struct diff_options *o,
2048                             int complete_rewrite)
2049{
2050        mmfile_t mf1, mf2;
2051        struct diffstat_file *data;
2052
2053        data = diffstat_add(diffstat, name_a, name_b);
2054
2055        if (!one || !two) {
2056                data->is_unmerged = 1;
2057                return;
2058        }
2059        if (complete_rewrite) {
2060                diff_populate_filespec(one, 0);
2061                diff_populate_filespec(two, 0);
2062                data->deleted = count_lines(one->data, one->size);
2063                data->added = count_lines(two->data, two->size);
2064                goto free_and_return;
2065        }
2066        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2067                die("unable to read files to diff");
2068
2069        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2070                data->is_binary = 1;
2071                data->added = mf2.size;
2072                data->deleted = mf1.size;
2073        } else {
2074                /* Crazy xdl interfaces.. */
2075                xpparam_t xpp;
2076                xdemitconf_t xecfg;
2077
2078                memset(&xpp, 0, sizeof(xpp));
2079                memset(&xecfg, 0, sizeof(xecfg));
2080                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
2081                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2082                              &xpp, &xecfg);
2083        }
2084
2085 free_and_return:
2086        diff_free_filespec_data(one);
2087        diff_free_filespec_data(two);
2088}
2089
2090static void builtin_checkdiff(const char *name_a, const char *name_b,
2091                              const char *attr_path,
2092                              struct diff_filespec *one,
2093                              struct diff_filespec *two,
2094                              struct diff_options *o)
2095{
2096        mmfile_t mf1, mf2;
2097        struct checkdiff_t data;
2098
2099        if (!two)
2100                return;
2101
2102        memset(&data, 0, sizeof(data));
2103        data.filename = name_b ? name_b : name_a;
2104        data.lineno = 0;
2105        data.o = o;
2106        data.ws_rule = whitespace_rule(attr_path);
2107        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2108
2109        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2110                die("unable to read files to diff");
2111
2112        /*
2113         * All the other codepaths check both sides, but not checking
2114         * the "old" side here is deliberate.  We are checking the newly
2115         * introduced changes, and as long as the "new" side is text, we
2116         * can and should check what it introduces.
2117         */
2118        if (diff_filespec_is_binary(two))
2119                goto free_and_return;
2120        else {
2121                /* Crazy xdl interfaces.. */
2122                xpparam_t xpp;
2123                xdemitconf_t xecfg;
2124
2125                memset(&xpp, 0, sizeof(xpp));
2126                memset(&xecfg, 0, sizeof(xecfg));
2127                xecfg.ctxlen = 1; /* at least one context line */
2128                xpp.flags = XDF_NEED_MINIMAL;
2129                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2130                              &xpp, &xecfg);
2131
2132                if (data.ws_rule & WS_BLANK_AT_EOF) {
2133                        struct emit_callback ecbdata;
2134                        int blank_at_eof;
2135
2136                        ecbdata.ws_rule = data.ws_rule;
2137                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2138                        blank_at_eof = ecbdata.blank_at_eof_in_preimage;
2139
2140                        if (blank_at_eof) {
2141                                static char *err;
2142                                if (!err)
2143                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2144                                fprintf(o->file, "%s:%d: %s.\n",
2145                                        data.filename, blank_at_eof, err);
2146                                data.status = 1; /* report errors */
2147                        }
2148                }
2149        }
2150 free_and_return:
2151        diff_free_filespec_data(one);
2152        diff_free_filespec_data(two);
2153        if (data.status)
2154                DIFF_OPT_SET(o, CHECK_FAILED);
2155}
2156
2157struct diff_filespec *alloc_filespec(const char *path)
2158{
2159        int namelen = strlen(path);
2160        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2161
2162        memset(spec, 0, sizeof(*spec));
2163        spec->path = (char *)(spec + 1);
2164        memcpy(spec->path, path, namelen+1);
2165        spec->count = 1;
2166        spec->is_binary = -1;
2167        return spec;
2168}
2169
2170void free_filespec(struct diff_filespec *spec)
2171{
2172        if (!--spec->count) {
2173                diff_free_filespec_data(spec);
2174                free(spec);
2175        }
2176}
2177
2178void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2179                   unsigned short mode)
2180{
2181        if (mode) {
2182                spec->mode = canon_mode(mode);
2183                hashcpy(spec->sha1, sha1);
2184                spec->sha1_valid = !is_null_sha1(sha1);
2185        }
2186}
2187
2188/*
2189 * Given a name and sha1 pair, if the index tells us the file in
2190 * the work tree has that object contents, return true, so that
2191 * prepare_temp_file() does not have to inflate and extract.
2192 */
2193static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2194{
2195        struct cache_entry *ce;
2196        struct stat st;
2197        int pos, len;
2198
2199        /*
2200         * We do not read the cache ourselves here, because the
2201         * benchmark with my previous version that always reads cache
2202         * shows that it makes things worse for diff-tree comparing
2203         * two linux-2.6 kernel trees in an already checked out work
2204         * tree.  This is because most diff-tree comparisons deal with
2205         * only a small number of files, while reading the cache is
2206         * expensive for a large project, and its cost outweighs the
2207         * savings we get by not inflating the object to a temporary
2208         * file.  Practically, this code only helps when we are used
2209         * by diff-cache --cached, which does read the cache before
2210         * calling us.
2211         */
2212        if (!active_cache)
2213                return 0;
2214
2215        /* We want to avoid the working directory if our caller
2216         * doesn't need the data in a normal file, this system
2217         * is rather slow with its stat/open/mmap/close syscalls,
2218         * and the object is contained in a pack file.  The pack
2219         * is probably already open and will be faster to obtain
2220         * the data through than the working directory.  Loose
2221         * objects however would tend to be slower as they need
2222         * to be individually opened and inflated.
2223         */
2224        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2225                return 0;
2226
2227        len = strlen(name);
2228        pos = cache_name_pos(name, len);
2229        if (pos < 0)
2230                return 0;
2231        ce = active_cache[pos];
2232
2233        /*
2234         * This is not the sha1 we are looking for, or
2235         * unreusable because it is not a regular file.
2236         */
2237        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2238                return 0;
2239
2240        /*
2241         * If ce is marked as "assume unchanged", there is no
2242         * guarantee that work tree matches what we are looking for.
2243         */
2244        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2245                return 0;
2246
2247        /*
2248         * If ce matches the file in the work tree, we can reuse it.
2249         */
2250        if (ce_uptodate(ce) ||
2251            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2252                return 1;
2253
2254        return 0;
2255}
2256
2257static int populate_from_stdin(struct diff_filespec *s)
2258{
2259        struct strbuf buf = STRBUF_INIT;
2260        size_t size = 0;
2261
2262        if (strbuf_read(&buf, 0, 0) < 0)
2263                return error("error while reading from stdin %s",
2264                                     strerror(errno));
2265
2266        s->should_munmap = 0;
2267        s->data = strbuf_detach(&buf, &size);
2268        s->size = size;
2269        s->should_free = 1;
2270        return 0;
2271}
2272
2273static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2274{
2275        int len;
2276        char *data = xmalloc(100), *dirty = "";
2277
2278        /* Are we looking at the work tree? */
2279        if (s->dirty_submodule)
2280                dirty = "-dirty";
2281
2282        len = snprintf(data, 100,
2283                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2284        s->data = data;
2285        s->size = len;
2286        s->should_free = 1;
2287        if (size_only) {
2288                s->data = NULL;
2289                free(data);
2290        }
2291        return 0;
2292}
2293
2294/*
2295 * While doing rename detection and pickaxe operation, we may need to
2296 * grab the data for the blob (or file) for our own in-core comparison.
2297 * diff_filespec has data and size fields for this purpose.
2298 */
2299int diff_populate_filespec(struct diff_filespec *s, int size_only)
2300{
2301        int err = 0;
2302        if (!DIFF_FILE_VALID(s))
2303                die("internal error: asking to populate invalid file.");
2304        if (S_ISDIR(s->mode))
2305                return -1;
2306
2307        if (s->data)
2308                return 0;
2309
2310        if (size_only && 0 < s->size)
2311                return 0;
2312
2313        if (S_ISGITLINK(s->mode))
2314                return diff_populate_gitlink(s, size_only);
2315
2316        if (!s->sha1_valid ||
2317            reuse_worktree_file(s->path, s->sha1, 0)) {
2318                struct strbuf buf = STRBUF_INIT;
2319                struct stat st;
2320                int fd;
2321
2322                if (!strcmp(s->path, "-"))
2323                        return populate_from_stdin(s);
2324
2325                if (lstat(s->path, &st) < 0) {
2326                        if (errno == ENOENT) {
2327                        err_empty:
2328                                err = -1;
2329                        empty:
2330                                s->data = (char *)"";
2331                                s->size = 0;
2332                                return err;
2333                        }
2334                }
2335                s->size = xsize_t(st.st_size);
2336                if (!s->size)
2337                        goto empty;
2338                if (S_ISLNK(st.st_mode)) {
2339                        struct strbuf sb = STRBUF_INIT;
2340
2341                        if (strbuf_readlink(&sb, s->path, s->size))
2342                                goto err_empty;
2343                        s->size = sb.len;
2344                        s->data = strbuf_detach(&sb, NULL);
2345                        s->should_free = 1;
2346                        return 0;
2347                }
2348                if (size_only)
2349                        return 0;
2350                fd = open(s->path, O_RDONLY);
2351                if (fd < 0)
2352                        goto err_empty;
2353                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2354                close(fd);
2355                s->should_munmap = 1;
2356
2357                /*
2358                 * Convert from working tree format to canonical git format
2359                 */
2360                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2361                        size_t size = 0;
2362                        munmap(s->data, s->size);
2363                        s->should_munmap = 0;
2364                        s->data = strbuf_detach(&buf, &size);
2365                        s->size = size;
2366                        s->should_free = 1;
2367                }
2368        }
2369        else {
2370                enum object_type type;
2371                if (size_only)
2372                        type = sha1_object_info(s->sha1, &s->size);
2373                else {
2374                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2375                        s->should_free = 1;
2376                }
2377        }
2378        return 0;
2379}
2380
2381void diff_free_filespec_blob(struct diff_filespec *s)
2382{
2383        if (s->should_free)
2384                free(s->data);
2385        else if (s->should_munmap)
2386                munmap(s->data, s->size);
2387
2388        if (s->should_free || s->should_munmap) {
2389                s->should_free = s->should_munmap = 0;
2390                s->data = NULL;
2391        }
2392}
2393
2394void diff_free_filespec_data(struct diff_filespec *s)
2395{
2396        diff_free_filespec_blob(s);
2397        free(s->cnt_data);
2398        s->cnt_data = NULL;
2399}
2400
2401static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2402                           void *blob,
2403                           unsigned long size,
2404                           const unsigned char *sha1,
2405                           int mode)
2406{
2407        int fd;
2408        struct strbuf buf = STRBUF_INIT;
2409        struct strbuf template = STRBUF_INIT;
2410        char *path_dup = xstrdup(path);
2411        const char *base = basename(path_dup);
2412
2413        /* Generate "XXXXXX_basename.ext" */
2414        strbuf_addstr(&template, "XXXXXX_");
2415        strbuf_addstr(&template, base);
2416
2417        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2418                        strlen(base) + 1);
2419        if (fd < 0)
2420                die_errno("unable to create temp-file");
2421        if (convert_to_working_tree(path,
2422                        (const char *)blob, (size_t)size, &buf)) {
2423                blob = buf.buf;
2424                size = buf.len;
2425        }
2426        if (write_in_full(fd, blob, size) != size)
2427                die_errno("unable to write temp-file");
2428        close(fd);
2429        temp->name = temp->tmp_path;
2430        strcpy(temp->hex, sha1_to_hex(sha1));
2431        temp->hex[40] = 0;
2432        sprintf(temp->mode, "%06o", mode);
2433        strbuf_release(&buf);
2434        strbuf_release(&template);
2435        free(path_dup);
2436}
2437
2438static struct diff_tempfile *prepare_temp_file(const char *name,
2439                struct diff_filespec *one)
2440{
2441        struct diff_tempfile *temp = claim_diff_tempfile();
2442
2443        if (!DIFF_FILE_VALID(one)) {
2444        not_a_valid_file:
2445                /* A '-' entry produces this for file-2, and
2446                 * a '+' entry produces this for file-1.
2447                 */
2448                temp->name = "/dev/null";
2449                strcpy(temp->hex, ".");
2450                strcpy(temp->mode, ".");
2451                return temp;
2452        }
2453
2454        if (!remove_tempfile_installed) {
2455                atexit(remove_tempfile);
2456                sigchain_push_common(remove_tempfile_on_signal);
2457                remove_tempfile_installed = 1;
2458        }
2459
2460        if (!one->sha1_valid ||
2461            reuse_worktree_file(name, one->sha1, 1)) {
2462                struct stat st;
2463                if (lstat(name, &st) < 0) {
2464                        if (errno == ENOENT)
2465                                goto not_a_valid_file;
2466                        die_errno("stat(%s)", name);
2467                }
2468                if (S_ISLNK(st.st_mode)) {
2469                        struct strbuf sb = STRBUF_INIT;
2470                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2471                                die_errno("readlink(%s)", name);
2472                        prep_temp_blob(name, temp, sb.buf, sb.len,
2473                                       (one->sha1_valid ?
2474                                        one->sha1 : null_sha1),
2475                                       (one->sha1_valid ?
2476                                        one->mode : S_IFLNK));
2477                        strbuf_release(&sb);
2478                }
2479                else {
2480                        /* we can borrow from the file in the work tree */
2481                        temp->name = name;
2482                        if (!one->sha1_valid)
2483                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2484                        else
2485                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2486                        /* Even though we may sometimes borrow the
2487                         * contents from the work tree, we always want
2488                         * one->mode.  mode is trustworthy even when
2489                         * !(one->sha1_valid), as long as
2490                         * DIFF_FILE_VALID(one).
2491                         */
2492                        sprintf(temp->mode, "%06o", one->mode);
2493                }
2494                return temp;
2495        }
2496        else {
2497                if (diff_populate_filespec(one, 0))
2498                        die("cannot read data blob for %s", one->path);
2499                prep_temp_blob(name, temp, one->data, one->size,
2500                               one->sha1, one->mode);
2501        }
2502        return temp;
2503}
2504
2505/* An external diff command takes:
2506 *
2507 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2508 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2509 *
2510 */
2511static void run_external_diff(const char *pgm,
2512                              const char *name,
2513                              const char *other,
2514                              struct diff_filespec *one,
2515                              struct diff_filespec *two,
2516                              const char *xfrm_msg,
2517                              int complete_rewrite)
2518{
2519        const char *spawn_arg[10];
2520        int retval;
2521        const char **arg = &spawn_arg[0];
2522
2523        if (one && two) {
2524                struct diff_tempfile *temp_one, *temp_two;
2525                const char *othername = (other ? other : name);
2526                temp_one = prepare_temp_file(name, one);
2527                temp_two = prepare_temp_file(othername, two);
2528                *arg++ = pgm;
2529                *arg++ = name;
2530                *arg++ = temp_one->name;
2531                *arg++ = temp_one->hex;
2532                *arg++ = temp_one->mode;
2533                *arg++ = temp_two->name;
2534                *arg++ = temp_two->hex;
2535                *arg++ = temp_two->mode;
2536                if (other) {
2537                        *arg++ = other;
2538                        *arg++ = xfrm_msg;
2539                }
2540        } else {
2541                *arg++ = pgm;
2542                *arg++ = name;
2543        }
2544        *arg = NULL;
2545        fflush(NULL);
2546        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2547        remove_tempfile();
2548        if (retval) {
2549                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2550                exit(1);
2551        }
2552}
2553
2554static int similarity_index(struct diff_filepair *p)
2555{
2556        return p->score * 100 / MAX_SCORE;
2557}
2558
2559static void fill_metainfo(struct strbuf *msg,
2560                          const char *name,
2561                          const char *other,
2562                          struct diff_filespec *one,
2563                          struct diff_filespec *two,
2564                          struct diff_options *o,
2565                          struct diff_filepair *p)
2566{
2567        struct strbuf *msgbuf;
2568        char *line_prefix = "";
2569
2570        if (o->output_prefix) {
2571                msgbuf = o->output_prefix(o, o->output_prefix_data);
2572                line_prefix = msgbuf->buf;
2573        }
2574
2575        strbuf_init(msg, PATH_MAX * 2 + 300);
2576        switch (p->status) {
2577        case DIFF_STATUS_COPIED:
2578                strbuf_addf(msg, "%ssimilarity index %d%%", line_prefix, similarity_index(p));
2579                strbuf_addf(msg, "\n%scopy from ", line_prefix);
2580                quote_c_style(name, msg, NULL, 0);
2581                strbuf_addf(msg, "\n%scopy to ", line_prefix);
2582                quote_c_style(other, msg, NULL, 0);
2583                strbuf_addch(msg, '\n');
2584                break;
2585        case DIFF_STATUS_RENAMED:
2586                strbuf_addf(msg, "%ssimilarity index %d%%", line_prefix, similarity_index(p));
2587                strbuf_addf(msg, "\n%srename from ", line_prefix);
2588                quote_c_style(name, msg, NULL, 0);
2589                strbuf_addf(msg, "\n%srename to ", line_prefix);
2590                quote_c_style(other, msg, NULL, 0);
2591                strbuf_addch(msg, '\n');
2592                break;
2593        case DIFF_STATUS_MODIFIED:
2594                if (p->score) {
2595                        strbuf_addf(msg, "%sdissimilarity index %d%%\n",
2596                                    line_prefix, similarity_index(p));
2597                        break;
2598                }
2599                /* fallthru */
2600        default:
2601                /* nothing */
2602                ;
2603        }
2604        if (one && two && hashcmp(one->sha1, two->sha1)) {
2605                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2606
2607                if (DIFF_OPT_TST(o, BINARY)) {
2608                        mmfile_t mf;
2609                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2610                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2611                                abbrev = 40;
2612                }
2613                strbuf_addf(msg, "%sindex %.*s..%.*s",
2614                            line_prefix, abbrev, sha1_to_hex(one->sha1),
2615                            abbrev, sha1_to_hex(two->sha1));
2616                if (one->mode == two->mode)
2617                        strbuf_addf(msg, " %06o", one->mode);
2618                strbuf_addch(msg, '\n');
2619        }
2620        if (msg->len)
2621                strbuf_setlen(msg, msg->len - 1);
2622}
2623
2624static void run_diff_cmd(const char *pgm,
2625                         const char *name,
2626                         const char *other,
2627                         const char *attr_path,
2628                         struct diff_filespec *one,
2629                         struct diff_filespec *two,
2630                         struct strbuf *msg,
2631                         struct diff_options *o,
2632                         struct diff_filepair *p)
2633{
2634        const char *xfrm_msg = NULL;
2635        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2636
2637        if (msg) {
2638                fill_metainfo(msg, name, other, one, two, o, p);
2639                xfrm_msg = msg->len ? msg->buf : NULL;
2640        }
2641
2642        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2643                pgm = NULL;
2644        else {
2645                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2646                if (drv && drv->external)
2647                        pgm = drv->external;
2648        }
2649
2650        if (pgm) {
2651                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2652                                  complete_rewrite);
2653                return;
2654        }
2655        if (one && two)
2656                builtin_diff(name, other ? other : name,
2657                             one, two, xfrm_msg, o, complete_rewrite);
2658        else
2659                fprintf(o->file, "* Unmerged path %s\n", name);
2660}
2661
2662static void diff_fill_sha1_info(struct diff_filespec *one)
2663{
2664        if (DIFF_FILE_VALID(one)) {
2665                if (!one->sha1_valid) {
2666                        struct stat st;
2667                        if (!strcmp(one->path, "-")) {
2668                                hashcpy(one->sha1, null_sha1);
2669                                return;
2670                        }
2671                        if (lstat(one->path, &st) < 0)
2672                                die_errno("stat '%s'", one->path);
2673                        if (index_path(one->sha1, one->path, &st, 0))
2674                                die("cannot hash %s", one->path);
2675                }
2676        }
2677        else
2678                hashclr(one->sha1);
2679}
2680
2681static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2682{
2683        /* Strip the prefix but do not molest /dev/null and absolute paths */
2684        if (*namep && **namep != '/')
2685                *namep += prefix_length;
2686        if (*otherp && **otherp != '/')
2687                *otherp += prefix_length;
2688}
2689
2690static void run_diff(struct diff_filepair *p, struct diff_options *o)
2691{
2692        const char *pgm = external_diff();
2693        struct strbuf msg;
2694        struct diff_filespec *one = p->one;
2695        struct diff_filespec *two = p->two;
2696        const char *name;
2697        const char *other;
2698        const char *attr_path;
2699
2700        name  = p->one->path;
2701        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2702        attr_path = name;
2703        if (o->prefix_length)
2704                strip_prefix(o->prefix_length, &name, &other);
2705
2706        if (DIFF_PAIR_UNMERGED(p)) {
2707                run_diff_cmd(pgm, name, NULL, attr_path,
2708                             NULL, NULL, NULL, o, p);
2709                return;
2710        }
2711
2712        diff_fill_sha1_info(one);
2713        diff_fill_sha1_info(two);
2714
2715        if (!pgm &&
2716            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2717            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2718                /*
2719                 * a filepair that changes between file and symlink
2720                 * needs to be split into deletion and creation.
2721                 */
2722                struct diff_filespec *null = alloc_filespec(two->path);
2723                run_diff_cmd(NULL, name, other, attr_path,
2724                             one, null, &msg, o, p);
2725                free(null);
2726                strbuf_release(&msg);
2727
2728                null = alloc_filespec(one->path);
2729                run_diff_cmd(NULL, name, other, attr_path,
2730                             null, two, &msg, o, p);
2731                free(null);
2732        }
2733        else
2734                run_diff_cmd(pgm, name, other, attr_path,
2735                             one, two, &msg, o, p);
2736
2737        strbuf_release(&msg);
2738}
2739
2740static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2741                         struct diffstat_t *diffstat)
2742{
2743        const char *name;
2744        const char *other;
2745        int complete_rewrite = 0;
2746
2747        if (DIFF_PAIR_UNMERGED(p)) {
2748                /* unmerged */
2749                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2750                return;
2751        }
2752
2753        name = p->one->path;
2754        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2755
2756        if (o->prefix_length)
2757                strip_prefix(o->prefix_length, &name, &other);
2758
2759        diff_fill_sha1_info(p->one);
2760        diff_fill_sha1_info(p->two);
2761
2762        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2763                complete_rewrite = 1;
2764        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2765}
2766
2767static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2768{
2769        const char *name;
2770        const char *other;
2771        const char *attr_path;
2772
2773        if (DIFF_PAIR_UNMERGED(p)) {
2774                /* unmerged */
2775                return;
2776        }
2777
2778        name = p->one->path;
2779        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2780        attr_path = other ? other : name;
2781
2782        if (o->prefix_length)
2783                strip_prefix(o->prefix_length, &name, &other);
2784
2785        diff_fill_sha1_info(p->one);
2786        diff_fill_sha1_info(p->two);
2787
2788        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2789}
2790
2791void diff_setup(struct diff_options *options)
2792{
2793        memset(options, 0, sizeof(*options));
2794        memset(&diff_queued_diff, 0, sizeof(diff_queued_diff));
2795
2796        options->file = stdout;
2797
2798        options->line_termination = '\n';
2799        options->break_opt = -1;
2800        options->rename_limit = -1;
2801        options->dirstat_percent = 3;
2802        options->context = 3;
2803
2804        options->change = diff_change;
2805        options->add_remove = diff_addremove;
2806        if (diff_use_color_default > 0)
2807                DIFF_OPT_SET(options, COLOR_DIFF);
2808        options->detect_rename = diff_detect_rename_default;
2809
2810        if (!diff_mnemonic_prefix) {
2811                options->a_prefix = "a/";
2812                options->b_prefix = "b/";
2813        }
2814}
2815
2816int diff_setup_done(struct diff_options *options)
2817{
2818        int count = 0;
2819
2820        if (options->output_format & DIFF_FORMAT_NAME)
2821                count++;
2822        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2823                count++;
2824        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2825                count++;
2826        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2827                count++;
2828        if (count > 1)
2829                die("--name-only, --name-status, --check and -s are mutually exclusive");
2830
2831        /*
2832         * Most of the time we can say "there are changes"
2833         * only by checking if there are changed paths, but
2834         * --ignore-whitespace* options force us to look
2835         * inside contents.
2836         */
2837
2838        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
2839            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
2840            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
2841                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
2842        else
2843                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
2844
2845        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2846                options->detect_rename = DIFF_DETECT_COPY;
2847
2848        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2849                options->prefix = NULL;
2850        if (options->prefix)
2851                options->prefix_length = strlen(options->prefix);
2852        else
2853                options->prefix_length = 0;
2854
2855        if (options->output_format & (DIFF_FORMAT_NAME |
2856                                      DIFF_FORMAT_NAME_STATUS |
2857                                      DIFF_FORMAT_CHECKDIFF |
2858                                      DIFF_FORMAT_NO_OUTPUT))
2859                options->output_format &= ~(DIFF_FORMAT_RAW |
2860                                            DIFF_FORMAT_NUMSTAT |
2861                                            DIFF_FORMAT_DIFFSTAT |
2862                                            DIFF_FORMAT_SHORTSTAT |
2863                                            DIFF_FORMAT_DIRSTAT |
2864                                            DIFF_FORMAT_SUMMARY |
2865                                            DIFF_FORMAT_PATCH);
2866
2867        /*
2868         * These cases always need recursive; we do not drop caller-supplied
2869         * recursive bits for other formats here.
2870         */
2871        if (options->output_format & (DIFF_FORMAT_PATCH |
2872                                      DIFF_FORMAT_NUMSTAT |
2873                                      DIFF_FORMAT_DIFFSTAT |
2874                                      DIFF_FORMAT_SHORTSTAT |
2875                                      DIFF_FORMAT_DIRSTAT |
2876                                      DIFF_FORMAT_SUMMARY |
2877                                      DIFF_FORMAT_CHECKDIFF))
2878                DIFF_OPT_SET(options, RECURSIVE);
2879        /*
2880         * Also pickaxe would not work very well if you do not say recursive
2881         */
2882        if (options->pickaxe)
2883                DIFF_OPT_SET(options, RECURSIVE);
2884        /*
2885         * When patches are generated, submodules diffed against the work tree
2886         * must be checked for dirtiness too so it can be shown in the output
2887         */
2888        if (options->output_format & DIFF_FORMAT_PATCH)
2889                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
2890
2891        if (options->detect_rename && options->rename_limit < 0)
2892                options->rename_limit = diff_rename_limit_default;
2893        if (options->setup & DIFF_SETUP_USE_CACHE) {
2894                if (!active_cache)
2895                        /* read-cache does not die even when it fails
2896                         * so it is safe for us to do this here.  Also
2897                         * it does not smudge active_cache or active_nr
2898                         * when it fails, so we do not have to worry about
2899                         * cleaning it up ourselves either.
2900                         */
2901                        read_cache();
2902        }
2903        if (options->abbrev <= 0 || 40 < options->abbrev)
2904                options->abbrev = 40; /* full */
2905
2906        /*
2907         * It does not make sense to show the first hit we happened
2908         * to have found.  It does not make sense not to return with
2909         * exit code in such a case either.
2910         */
2911        if (DIFF_OPT_TST(options, QUICK)) {
2912                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2913                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2914        }
2915
2916        return 0;
2917}
2918
2919static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2920{
2921        char c, *eq;
2922        int len;
2923
2924        if (*arg != '-')
2925                return 0;
2926        c = *++arg;
2927        if (!c)
2928                return 0;
2929        if (c == arg_short) {
2930                c = *++arg;
2931                if (!c)
2932                        return 1;
2933                if (val && isdigit(c)) {
2934                        char *end;
2935                        int n = strtoul(arg, &end, 10);
2936                        if (*end)
2937                                return 0;
2938                        *val = n;
2939                        return 1;
2940                }
2941                return 0;
2942        }
2943        if (c != '-')
2944                return 0;
2945        arg++;
2946        eq = strchr(arg, '=');
2947        if (eq)
2948                len = eq - arg;
2949        else
2950                len = strlen(arg);
2951        if (!len || strncmp(arg, arg_long, len))
2952                return 0;
2953        if (eq) {
2954                int n;
2955                char *end;
2956                if (!isdigit(*++eq))
2957                        return 0;
2958                n = strtoul(eq, &end, 10);
2959                if (*end)
2960                        return 0;
2961                *val = n;
2962        }
2963        return 1;
2964}
2965
2966static int diff_scoreopt_parse(const char *opt);
2967
2968int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2969{
2970        const char *arg = av[0];
2971
2972        /* Output format options */
2973        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
2974                options->output_format |= DIFF_FORMAT_PATCH;
2975        else if (opt_arg(arg, 'U', "unified", &options->context))
2976                options->output_format |= DIFF_FORMAT_PATCH;
2977        else if (!strcmp(arg, "--raw"))
2978                options->output_format |= DIFF_FORMAT_RAW;
2979        else if (!strcmp(arg, "--patch-with-raw"))
2980                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2981        else if (!strcmp(arg, "--numstat"))
2982                options->output_format |= DIFF_FORMAT_NUMSTAT;
2983        else if (!strcmp(arg, "--shortstat"))
2984                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2985        else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
2986                options->output_format |= DIFF_FORMAT_DIRSTAT;
2987        else if (!strcmp(arg, "--cumulative")) {
2988                options->output_format |= DIFF_FORMAT_DIRSTAT;
2989                DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
2990        } else if (opt_arg(arg, 0, "dirstat-by-file",
2991                           &options->dirstat_percent)) {
2992                options->output_format |= DIFF_FORMAT_DIRSTAT;
2993                DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
2994        }
2995        else if (!strcmp(arg, "--check"))
2996                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2997        else if (!strcmp(arg, "--summary"))
2998                options->output_format |= DIFF_FORMAT_SUMMARY;
2999        else if (!strcmp(arg, "--patch-with-stat"))
3000                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3001        else if (!strcmp(arg, "--name-only"))
3002                options->output_format |= DIFF_FORMAT_NAME;
3003        else if (!strcmp(arg, "--name-status"))
3004                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3005        else if (!strcmp(arg, "-s"))
3006                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3007        else if (!prefixcmp(arg, "--stat")) {
3008                char *end;
3009                int width = options->stat_width;
3010                int name_width = options->stat_name_width;
3011                arg += 6;
3012                end = (char *)arg;
3013
3014                switch (*arg) {
3015                case '-':
3016                        if (!prefixcmp(arg, "-width="))
3017                                width = strtoul(arg + 7, &end, 10);
3018                        else if (!prefixcmp(arg, "-name-width="))
3019                                name_width = strtoul(arg + 12, &end, 10);
3020                        break;
3021                case '=':
3022                        width = strtoul(arg+1, &end, 10);
3023                        if (*end == ',')
3024                                name_width = strtoul(end+1, &end, 10);
3025                }
3026
3027                /* Important! This checks all the error cases! */
3028                if (*end)
3029                        return 0;
3030                options->output_format |= DIFF_FORMAT_DIFFSTAT;
3031                options->stat_name_width = name_width;
3032                options->stat_width = width;
3033        }
3034
3035        /* renames options */
3036        else if (!prefixcmp(arg, "-B")) {
3037                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3038                        return -1;
3039        }
3040        else if (!prefixcmp(arg, "-M")) {
3041                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3042                        return -1;
3043                options->detect_rename = DIFF_DETECT_RENAME;
3044        }
3045        else if (!prefixcmp(arg, "-C")) {
3046                if (options->detect_rename == DIFF_DETECT_COPY)
3047                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3048                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3049                        return -1;
3050                options->detect_rename = DIFF_DETECT_COPY;
3051        }
3052        else if (!strcmp(arg, "--no-renames"))
3053                options->detect_rename = 0;
3054        else if (!strcmp(arg, "--relative"))
3055                DIFF_OPT_SET(options, RELATIVE_NAME);
3056        else if (!prefixcmp(arg, "--relative=")) {
3057                DIFF_OPT_SET(options, RELATIVE_NAME);
3058                options->prefix = arg + 11;
3059        }
3060
3061        /* xdiff options */
3062        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3063                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3064        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3065                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3066        else if (!strcmp(arg, "--ignore-space-at-eol"))
3067                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3068        else if (!strcmp(arg, "--patience"))
3069                DIFF_XDL_SET(options, PATIENCE_DIFF);
3070
3071        /* flags options */
3072        else if (!strcmp(arg, "--binary")) {
3073                options->output_format |= DIFF_FORMAT_PATCH;
3074                DIFF_OPT_SET(options, BINARY);
3075        }
3076        else if (!strcmp(arg, "--full-index"))
3077                DIFF_OPT_SET(options, FULL_INDEX);
3078        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3079                DIFF_OPT_SET(options, TEXT);
3080        else if (!strcmp(arg, "-R"))
3081                DIFF_OPT_SET(options, REVERSE_DIFF);
3082        else if (!strcmp(arg, "--find-copies-harder"))
3083                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3084        else if (!strcmp(arg, "--follow"))
3085                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3086        else if (!strcmp(arg, "--color"))
3087                DIFF_OPT_SET(options, COLOR_DIFF);
3088        else if (!prefixcmp(arg, "--color=")) {
3089                int value = git_config_colorbool(NULL, arg+8, -1);
3090                if (value == 0)
3091                        DIFF_OPT_CLR(options, COLOR_DIFF);
3092                else if (value > 0)
3093                        DIFF_OPT_SET(options, COLOR_DIFF);
3094                else
3095                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3096        }
3097        else if (!strcmp(arg, "--no-color"))
3098                DIFF_OPT_CLR(options, COLOR_DIFF);
3099        else if (!strcmp(arg, "--color-words")) {
3100                DIFF_OPT_SET(options, COLOR_DIFF);
3101                options->word_diff = DIFF_WORDS_COLOR;
3102        }
3103        else if (!prefixcmp(arg, "--color-words=")) {
3104                DIFF_OPT_SET(options, COLOR_DIFF);
3105                options->word_diff = DIFF_WORDS_COLOR;
3106                options->word_regex = arg + 14;
3107        }
3108        else if (!strcmp(arg, "--word-diff")) {
3109                if (options->word_diff == DIFF_WORDS_NONE)
3110                        options->word_diff = DIFF_WORDS_PLAIN;
3111        }
3112        else if (!prefixcmp(arg, "--word-diff=")) {
3113                const char *type = arg + 12;
3114                if (!strcmp(type, "plain"))
3115                        options->word_diff = DIFF_WORDS_PLAIN;
3116                else if (!strcmp(type, "color")) {
3117                        DIFF_OPT_SET(options, COLOR_DIFF);
3118                        options->word_diff = DIFF_WORDS_COLOR;
3119                }
3120                else if (!strcmp(type, "porcelain"))
3121                        options->word_diff = DIFF_WORDS_PORCELAIN;
3122                else if (!strcmp(type, "none"))
3123                        options->word_diff = DIFF_WORDS_NONE;
3124                else
3125                        die("bad --word-diff argument: %s", type);
3126        }
3127        else if (!prefixcmp(arg, "--word-diff-regex=")) {
3128                if (options->word_diff == DIFF_WORDS_NONE)
3129                        options->word_diff = DIFF_WORDS_PLAIN;
3130                options->word_regex = arg + 18;
3131        }
3132        else if (!strcmp(arg, "--exit-code"))
3133                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3134        else if (!strcmp(arg, "--quiet"))
3135                DIFF_OPT_SET(options, QUICK);
3136        else if (!strcmp(arg, "--ext-diff"))
3137                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3138        else if (!strcmp(arg, "--no-ext-diff"))
3139                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3140        else if (!strcmp(arg, "--textconv"))
3141                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3142        else if (!strcmp(arg, "--no-textconv"))
3143                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3144        else if (!strcmp(arg, "--ignore-submodules"))
3145                DIFF_OPT_SET(options, IGNORE_SUBMODULES);
3146        else if (!strcmp(arg, "--submodule"))
3147                DIFF_OPT_SET(options, SUBMODULE_LOG);
3148        else if (!prefixcmp(arg, "--submodule=")) {
3149                if (!strcmp(arg + 12, "log"))
3150                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3151        }
3152
3153        /* misc options */
3154        else if (!strcmp(arg, "-z"))
3155                options->line_termination = 0;
3156        else if (!prefixcmp(arg, "-l"))
3157                options->rename_limit = strtoul(arg+2, NULL, 10);
3158        else if (!prefixcmp(arg, "-S"))
3159                options->pickaxe = arg + 2;
3160        else if (!strcmp(arg, "--pickaxe-all"))
3161                options->pickaxe_opts = DIFF_PICKAXE_ALL;
3162        else if (!strcmp(arg, "--pickaxe-regex"))
3163                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
3164        else if (!prefixcmp(arg, "-O"))
3165                options->orderfile = arg + 2;
3166        else if (!prefixcmp(arg, "--diff-filter="))
3167                options->filter = arg + 14;
3168        else if (!strcmp(arg, "--abbrev"))
3169                options->abbrev = DEFAULT_ABBREV;
3170        else if (!prefixcmp(arg, "--abbrev=")) {
3171                options->abbrev = strtoul(arg + 9, NULL, 10);
3172                if (options->abbrev < MINIMUM_ABBREV)
3173                        options->abbrev = MINIMUM_ABBREV;
3174                else if (40 < options->abbrev)
3175                        options->abbrev = 40;
3176        }
3177        else if (!prefixcmp(arg, "--src-prefix="))
3178                options->a_prefix = arg + 13;
3179        else if (!prefixcmp(arg, "--dst-prefix="))
3180                options->b_prefix = arg + 13;
3181        else if (!strcmp(arg, "--no-prefix"))
3182                options->a_prefix = options->b_prefix = "";
3183        else if (opt_arg(arg, '\0', "inter-hunk-context",
3184                         &options->interhunkcontext))
3185                ;
3186        else if (!prefixcmp(arg, "--output=")) {
3187                options->file = fopen(arg + strlen("--output="), "w");
3188                if (!options->file)
3189                        die_errno("Could not open '%s'", arg + strlen("--output="));
3190                options->close_file = 1;
3191        } else
3192                return 0;
3193        return 1;
3194}
3195
3196static int parse_num(const char **cp_p)
3197{
3198        unsigned long num, scale;
3199        int ch, dot;
3200        const char *cp = *cp_p;
3201
3202        num = 0;
3203        scale = 1;
3204        dot = 0;
3205        for (;;) {
3206                ch = *cp;
3207                if ( !dot && ch == '.' ) {
3208                        scale = 1;
3209                        dot = 1;
3210                } else if ( ch == '%' ) {
3211                        scale = dot ? scale*100 : 100;
3212                        cp++;   /* % is always at the end */
3213                        break;
3214                } else if ( ch >= '0' && ch <= '9' ) {
3215                        if ( scale < 100000 ) {
3216                                scale *= 10;
3217                                num = (num*10) + (ch-'0');
3218                        }
3219                } else {
3220                        break;
3221                }
3222                cp++;
3223        }
3224        *cp_p = cp;
3225
3226        /* user says num divided by scale and we say internally that
3227         * is MAX_SCORE * num / scale.
3228         */
3229        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3230}
3231
3232static int diff_scoreopt_parse(const char *opt)
3233{
3234        int opt1, opt2, cmd;
3235
3236        if (*opt++ != '-')
3237                return -1;
3238        cmd = *opt++;
3239        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3240                return -1; /* that is not a -M, -C nor -B option */
3241
3242        opt1 = parse_num(&opt);
3243        if (cmd != 'B')
3244                opt2 = 0;
3245        else {
3246                if (*opt == 0)
3247                        opt2 = 0;
3248                else if (*opt != '/')
3249                        return -1; /* we expect -B80/99 or -B80 */
3250                else {
3251                        opt++;
3252                        opt2 = parse_num(&opt);
3253                }
3254        }
3255        if (*opt != 0)
3256                return -1;
3257        return opt1 | (opt2 << 16);
3258}
3259
3260struct diff_queue_struct diff_queued_diff;
3261
3262void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3263{
3264        if (queue->alloc <= queue->nr) {
3265                queue->alloc = alloc_nr(queue->alloc);
3266                queue->queue = xrealloc(queue->queue,
3267                                        sizeof(dp) * queue->alloc);
3268        }
3269        queue->queue[queue->nr++] = dp;
3270}
3271
3272struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3273                                 struct diff_filespec *one,
3274                                 struct diff_filespec *two)
3275{
3276        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3277        dp->one = one;
3278        dp->two = two;
3279        if (queue)
3280                diff_q(queue, dp);
3281        return dp;
3282}
3283
3284void diff_free_filepair(struct diff_filepair *p)
3285{
3286        free_filespec(p->one);
3287        free_filespec(p->two);
3288        free(p);
3289}
3290
3291/* This is different from find_unique_abbrev() in that
3292 * it stuffs the result with dots for alignment.
3293 */
3294const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3295{
3296        int abblen;
3297        const char *abbrev;
3298        if (len == 40)
3299                return sha1_to_hex(sha1);
3300
3301        abbrev = find_unique_abbrev(sha1, len);
3302        abblen = strlen(abbrev);
3303        if (abblen < 37) {
3304                static char hex[41];
3305                if (len < abblen && abblen <= len + 2)
3306                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3307                else
3308                        sprintf(hex, "%s...", abbrev);
3309                return hex;
3310        }
3311        return sha1_to_hex(sha1);
3312}
3313
3314static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3315{
3316        int line_termination = opt->line_termination;
3317        int inter_name_termination = line_termination ? '\t' : '\0';
3318        if (opt->output_prefix) {
3319                struct strbuf *msg = NULL;
3320                msg = opt->output_prefix(opt, opt->output_prefix_data);
3321                fprintf(opt->file, "%s", msg->buf);
3322        }
3323
3324        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3325                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3326                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3327                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3328        }
3329        if (p->score) {
3330                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3331                        inter_name_termination);
3332        } else {
3333                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3334        }
3335
3336        if (p->status == DIFF_STATUS_COPIED ||
3337            p->status == DIFF_STATUS_RENAMED) {
3338                const char *name_a, *name_b;
3339                name_a = p->one->path;
3340                name_b = p->two->path;
3341                strip_prefix(opt->prefix_length, &name_a, &name_b);
3342                write_name_quoted(name_a, opt->file, inter_name_termination);
3343                write_name_quoted(name_b, opt->file, line_termination);
3344        } else {
3345                const char *name_a, *name_b;
3346                name_a = p->one->mode ? p->one->path : p->two->path;
3347                name_b = NULL;
3348                strip_prefix(opt->prefix_length, &name_a, &name_b);
3349                write_name_quoted(name_a, opt->file, line_termination);
3350        }
3351}
3352
3353int diff_unmodified_pair(struct diff_filepair *p)
3354{
3355        /* This function is written stricter than necessary to support
3356         * the currently implemented transformers, but the idea is to
3357         * let transformers to produce diff_filepairs any way they want,
3358         * and filter and clean them up here before producing the output.
3359         */
3360        struct diff_filespec *one = p->one, *two = p->two;
3361
3362        if (DIFF_PAIR_UNMERGED(p))
3363                return 0; /* unmerged is interesting */
3364
3365        /* deletion, addition, mode or type change
3366         * and rename are all interesting.
3367         */
3368        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3369            DIFF_PAIR_MODE_CHANGED(p) ||
3370            strcmp(one->path, two->path))
3371                return 0;
3372
3373        /* both are valid and point at the same path.  that is, we are
3374         * dealing with a change.
3375         */
3376        if (one->sha1_valid && two->sha1_valid &&
3377            !hashcmp(one->sha1, two->sha1) &&
3378            !one->dirty_submodule && !two->dirty_submodule)
3379                return 1; /* no change */
3380        if (!one->sha1_valid && !two->sha1_valid)
3381                return 1; /* both look at the same file on the filesystem. */
3382        return 0;
3383}
3384
3385static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3386{
3387        if (diff_unmodified_pair(p))
3388                return;
3389
3390        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3391            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3392                return; /* no tree diffs in patch format */
3393
3394        run_diff(p, o);
3395}
3396
3397static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3398                            struct diffstat_t *diffstat)
3399{
3400        if (diff_unmodified_pair(p))
3401                return;
3402
3403        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3404            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3405                return; /* no tree diffs in patch format */
3406
3407        run_diffstat(p, o, diffstat);
3408}
3409
3410static void diff_flush_checkdiff(struct diff_filepair *p,
3411                struct diff_options *o)
3412{
3413        if (diff_unmodified_pair(p))
3414                return;
3415
3416        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3417            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3418                return; /* no tree diffs in patch format */
3419
3420        run_checkdiff(p, o);
3421}
3422
3423int diff_queue_is_empty(void)
3424{
3425        struct diff_queue_struct *q = &diff_queued_diff;
3426        int i;
3427        for (i = 0; i < q->nr; i++)
3428                if (!diff_unmodified_pair(q->queue[i]))
3429                        return 0;
3430        return 1;
3431}
3432
3433#if DIFF_DEBUG
3434void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3435{
3436        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3437                x, one ? one : "",
3438                s->path,
3439                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3440                s->mode,
3441                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3442        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3443                x, one ? one : "",
3444                s->size, s->xfrm_flags);
3445}
3446
3447void diff_debug_filepair(const struct diff_filepair *p, int i)
3448{
3449        diff_debug_filespec(p->one, i, "one");
3450        diff_debug_filespec(p->two, i, "two");
3451        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3452                p->score, p->status ? p->status : '?',
3453                p->one->rename_used, p->broken_pair);
3454}
3455
3456void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3457{
3458        int i;
3459        if (msg)
3460                fprintf(stderr, "%s\n", msg);
3461        fprintf(stderr, "q->nr = %d\n", q->nr);
3462        for (i = 0; i < q->nr; i++) {
3463                struct diff_filepair *p = q->queue[i];
3464                diff_debug_filepair(p, i);
3465        }
3466}
3467#endif
3468
3469static void diff_resolve_rename_copy(void)
3470{
3471        int i;
3472        struct diff_filepair *p;
3473        struct diff_queue_struct *q = &diff_queued_diff;
3474
3475        diff_debug_queue("resolve-rename-copy", q);
3476
3477        for (i = 0; i < q->nr; i++) {
3478                p = q->queue[i];
3479                p->status = 0; /* undecided */
3480                if (DIFF_PAIR_UNMERGED(p))
3481                        p->status = DIFF_STATUS_UNMERGED;
3482                else if (!DIFF_FILE_VALID(p->one))
3483                        p->status = DIFF_STATUS_ADDED;
3484                else if (!DIFF_FILE_VALID(p->two))
3485                        p->status = DIFF_STATUS_DELETED;
3486                else if (DIFF_PAIR_TYPE_CHANGED(p))
3487                        p->status = DIFF_STATUS_TYPE_CHANGED;
3488
3489                /* from this point on, we are dealing with a pair
3490                 * whose both sides are valid and of the same type, i.e.
3491                 * either in-place edit or rename/copy edit.
3492                 */
3493                else if (DIFF_PAIR_RENAME(p)) {
3494                        /*
3495                         * A rename might have re-connected a broken
3496                         * pair up, causing the pathnames to be the
3497                         * same again. If so, that's not a rename at
3498                         * all, just a modification..
3499                         *
3500                         * Otherwise, see if this source was used for
3501                         * multiple renames, in which case we decrement
3502                         * the count, and call it a copy.
3503                         */
3504                        if (!strcmp(p->one->path, p->two->path))
3505                                p->status = DIFF_STATUS_MODIFIED;
3506                        else if (--p->one->rename_used > 0)
3507                                p->status = DIFF_STATUS_COPIED;
3508                        else
3509                                p->status = DIFF_STATUS_RENAMED;
3510                }
3511                else if (hashcmp(p->one->sha1, p->two->sha1) ||
3512                         p->one->mode != p->two->mode ||
3513                         p->one->dirty_submodule ||
3514                         p->two->dirty_submodule ||
3515                         is_null_sha1(p->one->sha1))
3516                        p->status = DIFF_STATUS_MODIFIED;
3517                else {
3518                        /* This is a "no-change" entry and should not
3519                         * happen anymore, but prepare for broken callers.
3520                         */
3521                        error("feeding unmodified %s to diffcore",
3522                              p->one->path);
3523                        p->status = DIFF_STATUS_UNKNOWN;
3524                }
3525        }
3526        diff_debug_queue("resolve-rename-copy done", q);
3527}
3528
3529static int check_pair_status(struct diff_filepair *p)
3530{
3531        switch (p->status) {
3532        case DIFF_STATUS_UNKNOWN:
3533                return 0;
3534        case 0:
3535                die("internal error in diff-resolve-rename-copy");
3536        default:
3537                return 1;
3538        }
3539}
3540
3541static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3542{
3543        int fmt = opt->output_format;
3544
3545        if (fmt & DIFF_FORMAT_CHECKDIFF)
3546                diff_flush_checkdiff(p, opt);
3547        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3548                diff_flush_raw(p, opt);
3549        else if (fmt & DIFF_FORMAT_NAME) {
3550                const char *name_a, *name_b;
3551                name_a = p->two->path;
3552                name_b = NULL;
3553                strip_prefix(opt->prefix_length, &name_a, &name_b);
3554                write_name_quoted(name_a, opt->file, opt->line_termination);
3555        }
3556}
3557
3558static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3559{
3560        if (fs->mode)
3561                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3562        else
3563                fprintf(file, " %s ", newdelete);
3564        write_name_quoted(fs->path, file, '\n');
3565}
3566
3567
3568static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
3569                const char *line_prefix)
3570{
3571        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3572                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
3573                        p->two->mode, show_name ? ' ' : '\n');
3574                if (show_name) {
3575                        write_name_quoted(p->two->path, file, '\n');
3576                }
3577        }
3578}
3579
3580static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
3581                        const char *line_prefix)
3582{
3583        char *names = pprint_rename(p->one->path, p->two->path);
3584
3585        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3586        free(names);
3587        show_mode_change(file, p, 0, line_prefix);
3588}
3589
3590static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
3591{
3592        FILE *file = opt->file;
3593        char *line_prefix = "";
3594
3595        if (opt->output_prefix) {
3596                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
3597                line_prefix = buf->buf;
3598        }
3599
3600        switch(p->status) {
3601        case DIFF_STATUS_DELETED:
3602                fputs(line_prefix, file);
3603                show_file_mode_name(file, "delete", p->one);
3604                break;
3605        case DIFF_STATUS_ADDED:
3606                fputs(line_prefix, file);
3607                show_file_mode_name(file, "create", p->two);
3608                break;
3609        case DIFF_STATUS_COPIED:
3610                fputs(line_prefix, file);
3611                show_rename_copy(file, "copy", p, line_prefix);
3612                break;
3613        case DIFF_STATUS_RENAMED:
3614                fputs(line_prefix, file);
3615                show_rename_copy(file, "rename", p, line_prefix);
3616                break;
3617        default:
3618                if (p->score) {
3619                        fprintf(file, "%s rewrite ", line_prefix);
3620                        write_name_quoted(p->two->path, file, ' ');
3621                        fprintf(file, "(%d%%)\n", similarity_index(p));
3622                }
3623                show_mode_change(file, p, !p->score, line_prefix);
3624                break;
3625        }
3626}
3627
3628struct patch_id_t {
3629        git_SHA_CTX *ctx;
3630        int patchlen;
3631};
3632
3633static int remove_space(char *line, int len)
3634{
3635        int i;
3636        char *dst = line;
3637        unsigned char c;
3638
3639        for (i = 0; i < len; i++)
3640                if (!isspace((c = line[i])))
3641                        *dst++ = c;
3642
3643        return dst - line;
3644}
3645
3646static void patch_id_consume(void *priv, char *line, unsigned long len)
3647{
3648        struct patch_id_t *data = priv;
3649        int new_len;
3650
3651        /* Ignore line numbers when computing the SHA1 of the patch */
3652        if (!prefixcmp(line, "@@ -"))
3653                return;
3654
3655        new_len = remove_space(line, len);
3656
3657        git_SHA1_Update(data->ctx, line, new_len);
3658        data->patchlen += new_len;
3659}
3660
3661/* returns 0 upon success, and writes result into sha1 */
3662static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
3663{
3664        struct diff_queue_struct *q = &diff_queued_diff;
3665        int i;
3666        git_SHA_CTX ctx;
3667        struct patch_id_t data;
3668        char buffer[PATH_MAX * 4 + 20];
3669
3670        git_SHA1_Init(&ctx);
3671        memset(&data, 0, sizeof(struct patch_id_t));
3672        data.ctx = &ctx;
3673
3674        for (i = 0; i < q->nr; i++) {
3675                xpparam_t xpp;
3676                xdemitconf_t xecfg;
3677                mmfile_t mf1, mf2;
3678                struct diff_filepair *p = q->queue[i];
3679                int len1, len2;
3680
3681                memset(&xpp, 0, sizeof(xpp));
3682                memset(&xecfg, 0, sizeof(xecfg));
3683                if (p->status == 0)
3684                        return error("internal diff status error");
3685                if (p->status == DIFF_STATUS_UNKNOWN)
3686                        continue;
3687                if (diff_unmodified_pair(p))
3688                        continue;
3689                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3690                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3691                        continue;
3692                if (DIFF_PAIR_UNMERGED(p))
3693                        continue;
3694
3695                diff_fill_sha1_info(p->one);
3696                diff_fill_sha1_info(p->two);
3697                if (fill_mmfile(&mf1, p->one) < 0 ||
3698                                fill_mmfile(&mf2, p->two) < 0)
3699                        return error("unable to read files to diff");
3700
3701                len1 = remove_space(p->one->path, strlen(p->one->path));
3702                len2 = remove_space(p->two->path, strlen(p->two->path));
3703                if (p->one->mode == 0)
3704                        len1 = snprintf(buffer, sizeof(buffer),
3705                                        "diff--gita/%.*sb/%.*s"
3706                                        "newfilemode%06o"
3707                                        "---/dev/null"
3708                                        "+++b/%.*s",
3709                                        len1, p->one->path,
3710                                        len2, p->two->path,
3711                                        p->two->mode,
3712                                        len2, p->two->path);
3713                else if (p->two->mode == 0)
3714                        len1 = snprintf(buffer, sizeof(buffer),
3715                                        "diff--gita/%.*sb/%.*s"
3716                                        "deletedfilemode%06o"
3717                                        "---a/%.*s"
3718                                        "+++/dev/null",
3719                                        len1, p->one->path,
3720                                        len2, p->two->path,
3721                                        p->one->mode,
3722                                        len1, p->one->path);
3723                else
3724                        len1 = snprintf(buffer, sizeof(buffer),
3725                                        "diff--gita/%.*sb/%.*s"
3726                                        "---a/%.*s"
3727                                        "+++b/%.*s",
3728                                        len1, p->one->path,
3729                                        len2, p->two->path,
3730                                        len1, p->one->path,
3731                                        len2, p->two->path);
3732                git_SHA1_Update(&ctx, buffer, len1);
3733
3734                xpp.flags = XDF_NEED_MINIMAL;
3735                xecfg.ctxlen = 3;
3736                xecfg.flags = XDL_EMIT_FUNCNAMES;
3737                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
3738                              &xpp, &xecfg);
3739        }
3740
3741        git_SHA1_Final(sha1, &ctx);
3742        return 0;
3743}
3744
3745int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3746{
3747        struct diff_queue_struct *q = &diff_queued_diff;
3748        int i;
3749        int result = diff_get_patch_id(options, sha1);
3750
3751        for (i = 0; i < q->nr; i++)
3752                diff_free_filepair(q->queue[i]);
3753
3754        free(q->queue);
3755        DIFF_QUEUE_CLEAR(q);
3756
3757        return result;
3758}
3759
3760static int is_summary_empty(const struct diff_queue_struct *q)
3761{
3762        int i;
3763
3764        for (i = 0; i < q->nr; i++) {
3765                const struct diff_filepair *p = q->queue[i];
3766
3767                switch (p->status) {
3768                case DIFF_STATUS_DELETED:
3769                case DIFF_STATUS_ADDED:
3770                case DIFF_STATUS_COPIED:
3771                case DIFF_STATUS_RENAMED:
3772                        return 0;
3773                default:
3774                        if (p->score)
3775                                return 0;
3776                        if (p->one->mode && p->two->mode &&
3777                            p->one->mode != p->two->mode)
3778                                return 0;
3779                        break;
3780                }
3781        }
3782        return 1;
3783}
3784
3785void diff_flush(struct diff_options *options)
3786{
3787        struct diff_queue_struct *q = &diff_queued_diff;
3788        int i, output_format = options->output_format;
3789        int separator = 0;
3790
3791        /*
3792         * Order: raw, stat, summary, patch
3793         * or:    name/name-status/checkdiff (other bits clear)
3794         */
3795        if (!q->nr)
3796                goto free_queue;
3797
3798        if (output_format & (DIFF_FORMAT_RAW |
3799                             DIFF_FORMAT_NAME |
3800                             DIFF_FORMAT_NAME_STATUS |
3801                             DIFF_FORMAT_CHECKDIFF)) {
3802                for (i = 0; i < q->nr; i++) {
3803                        struct diff_filepair *p = q->queue[i];
3804                        if (check_pair_status(p))
3805                                flush_one_pair(p, options);
3806                }
3807                separator++;
3808        }
3809
3810        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
3811                struct diffstat_t diffstat;
3812
3813                memset(&diffstat, 0, sizeof(struct diffstat_t));
3814                for (i = 0; i < q->nr; i++) {
3815                        struct diff_filepair *p = q->queue[i];
3816                        if (check_pair_status(p))
3817                                diff_flush_stat(p, options, &diffstat);
3818                }
3819                if (output_format & DIFF_FORMAT_NUMSTAT)
3820                        show_numstat(&diffstat, options);
3821                if (output_format & DIFF_FORMAT_DIFFSTAT)
3822                        show_stats(&diffstat, options);
3823                if (output_format & DIFF_FORMAT_SHORTSTAT)
3824                        show_shortstats(&diffstat, options);
3825                free_diffstat_info(&diffstat);
3826                separator++;
3827        }
3828        if (output_format & DIFF_FORMAT_DIRSTAT)
3829                show_dirstat(options);
3830
3831        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
3832                for (i = 0; i < q->nr; i++) {
3833                        diff_summary(options, q->queue[i]);
3834                }
3835                separator++;
3836        }
3837
3838        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
3839            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
3840            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
3841                /*
3842                 * run diff_flush_patch for the exit status. setting
3843                 * options->file to /dev/null should be safe, becaue we
3844                 * aren't supposed to produce any output anyway.
3845                 */
3846                if (options->close_file)
3847                        fclose(options->file);
3848                options->file = fopen("/dev/null", "w");
3849                if (!options->file)
3850                        die_errno("Could not open /dev/null");
3851                options->close_file = 1;
3852                for (i = 0; i < q->nr; i++) {
3853                        struct diff_filepair *p = q->queue[i];
3854                        if (check_pair_status(p))
3855                                diff_flush_patch(p, options);
3856                        if (options->found_changes)
3857                                break;
3858                }
3859        }
3860
3861        if (output_format & DIFF_FORMAT_PATCH) {
3862                if (separator) {
3863                        putc(options->line_termination, options->file);
3864                        if (options->stat_sep) {
3865                                /* attach patch instead of inline */
3866                                fputs(options->stat_sep, options->file);
3867                        }
3868                }
3869
3870                for (i = 0; i < q->nr; i++) {
3871                        struct diff_filepair *p = q->queue[i];
3872                        if (check_pair_status(p))
3873                                diff_flush_patch(p, options);
3874                }
3875        }
3876
3877        if (output_format & DIFF_FORMAT_CALLBACK)
3878                options->format_callback(q, options, options->format_callback_data);
3879
3880        for (i = 0; i < q->nr; i++)
3881                diff_free_filepair(q->queue[i]);
3882free_queue:
3883        free(q->queue);
3884        DIFF_QUEUE_CLEAR(q);
3885        if (options->close_file)
3886                fclose(options->file);
3887
3888        /*
3889         * Report the content-level differences with HAS_CHANGES;
3890         * diff_addremove/diff_change does not set the bit when
3891         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
3892         */
3893        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
3894                if (options->found_changes)
3895                        DIFF_OPT_SET(options, HAS_CHANGES);
3896                else
3897                        DIFF_OPT_CLR(options, HAS_CHANGES);
3898        }
3899}
3900
3901static void diffcore_apply_filter(const char *filter)
3902{
3903        int i;
3904        struct diff_queue_struct *q = &diff_queued_diff;
3905        struct diff_queue_struct outq;
3906        DIFF_QUEUE_CLEAR(&outq);
3907
3908        if (!filter)
3909                return;
3910
3911        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3912                int found;
3913                for (i = found = 0; !found && i < q->nr; i++) {
3914                        struct diff_filepair *p = q->queue[i];
3915                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3916                             ((p->score &&
3917                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3918                              (!p->score &&
3919                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3920                            ((p->status != DIFF_STATUS_MODIFIED) &&
3921                             strchr(filter, p->status)))
3922                                found++;
3923                }
3924                if (found)
3925                        return;
3926
3927                /* otherwise we will clear the whole queue
3928                 * by copying the empty outq at the end of this
3929                 * function, but first clear the current entries
3930                 * in the queue.
3931                 */
3932                for (i = 0; i < q->nr; i++)
3933                        diff_free_filepair(q->queue[i]);
3934        }
3935        else {
3936                /* Only the matching ones */
3937                for (i = 0; i < q->nr; i++) {
3938                        struct diff_filepair *p = q->queue[i];
3939
3940                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3941                             ((p->score &&
3942                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3943                              (!p->score &&
3944                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3945                            ((p->status != DIFF_STATUS_MODIFIED) &&
3946                             strchr(filter, p->status)))
3947                                diff_q(&outq, p);
3948                        else
3949                                diff_free_filepair(p);
3950                }
3951        }
3952        free(q->queue);
3953        *q = outq;
3954}
3955
3956/* Check whether two filespecs with the same mode and size are identical */
3957static int diff_filespec_is_identical(struct diff_filespec *one,
3958                                      struct diff_filespec *two)
3959{
3960        if (S_ISGITLINK(one->mode))
3961                return 0;
3962        if (diff_populate_filespec(one, 0))
3963                return 0;
3964        if (diff_populate_filespec(two, 0))
3965                return 0;
3966        return !memcmp(one->data, two->data, one->size);
3967}
3968
3969static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3970{
3971        int i;
3972        struct diff_queue_struct *q = &diff_queued_diff;
3973        struct diff_queue_struct outq;
3974        DIFF_QUEUE_CLEAR(&outq);
3975
3976        for (i = 0; i < q->nr; i++) {
3977                struct diff_filepair *p = q->queue[i];
3978
3979                /*
3980                 * 1. Entries that come from stat info dirtiness
3981                 *    always have both sides (iow, not create/delete),
3982                 *    one side of the object name is unknown, with
3983                 *    the same mode and size.  Keep the ones that
3984                 *    do not match these criteria.  They have real
3985                 *    differences.
3986                 *
3987                 * 2. At this point, the file is known to be modified,
3988                 *    with the same mode and size, and the object
3989                 *    name of one side is unknown.  Need to inspect
3990                 *    the identical contents.
3991                 */
3992                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3993                    !DIFF_FILE_VALID(p->two) ||
3994                    (p->one->sha1_valid && p->two->sha1_valid) ||
3995                    (p->one->mode != p->two->mode) ||
3996                    diff_populate_filespec(p->one, 1) ||
3997                    diff_populate_filespec(p->two, 1) ||
3998                    (p->one->size != p->two->size) ||
3999                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4000                        diff_q(&outq, p);
4001                else {
4002                        /*
4003                         * The caller can subtract 1 from skip_stat_unmatch
4004                         * to determine how many paths were dirty only
4005                         * due to stat info mismatch.
4006                         */
4007                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4008                                diffopt->skip_stat_unmatch++;
4009                        diff_free_filepair(p);
4010                }
4011        }
4012        free(q->queue);
4013        *q = outq;
4014}
4015
4016static int diffnamecmp(const void *a_, const void *b_)
4017{
4018        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4019        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4020        const char *name_a, *name_b;
4021
4022        name_a = a->one ? a->one->path : a->two->path;
4023        name_b = b->one ? b->one->path : b->two->path;
4024        return strcmp(name_a, name_b);
4025}
4026
4027void diffcore_fix_diff_index(struct diff_options *options)
4028{
4029        struct diff_queue_struct *q = &diff_queued_diff;
4030        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4031}
4032
4033void diffcore_std(struct diff_options *options)
4034{
4035        /* We never run this function more than one time, because the
4036         * rename/copy detection logic can only run once.
4037         */
4038        if (diff_queued_diff.run)
4039                return;
4040
4041        if (options->skip_stat_unmatch)
4042                diffcore_skip_stat_unmatch(options);
4043        if (options->break_opt != -1)
4044                diffcore_break(options->break_opt);
4045        if (options->detect_rename)
4046                diffcore_rename(options);
4047        if (options->break_opt != -1)
4048                diffcore_merge_broken();
4049        if (options->pickaxe)
4050                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
4051        if (options->orderfile)
4052                diffcore_order(options->orderfile);
4053        diff_resolve_rename_copy();
4054        diffcore_apply_filter(options->filter);
4055
4056        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4057                DIFF_OPT_SET(options, HAS_CHANGES);
4058        else
4059                DIFF_OPT_CLR(options, HAS_CHANGES);
4060
4061        diff_queued_diff.run = 1;
4062}
4063
4064int diff_result_code(struct diff_options *opt, int status)
4065{
4066        int result = 0;
4067        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4068            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4069                return status;
4070        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4071            DIFF_OPT_TST(opt, HAS_CHANGES))
4072                result |= 01;
4073        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4074            DIFF_OPT_TST(opt, CHECK_FAILED))
4075                result |= 02;
4076        return result;
4077}
4078
4079void diff_addremove(struct diff_options *options,
4080                    int addremove, unsigned mode,
4081                    const unsigned char *sha1,
4082                    const char *concatpath, unsigned dirty_submodule)
4083{
4084        struct diff_filespec *one, *two;
4085
4086        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
4087                return;
4088
4089        /* This may look odd, but it is a preparation for
4090         * feeding "there are unchanged files which should
4091         * not produce diffs, but when you are doing copy
4092         * detection you would need them, so here they are"
4093         * entries to the diff-core.  They will be prefixed
4094         * with something like '=' or '*' (I haven't decided
4095         * which but should not make any difference).
4096         * Feeding the same new and old to diff_change()
4097         * also has the same effect.
4098         * Before the final output happens, they are pruned after
4099         * merged into rename/copy pairs as appropriate.
4100         */
4101        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4102                addremove = (addremove == '+' ? '-' :
4103                             addremove == '-' ? '+' : addremove);
4104
4105        if (options->prefix &&
4106            strncmp(concatpath, options->prefix, options->prefix_length))
4107                return;
4108
4109        one = alloc_filespec(concatpath);
4110        two = alloc_filespec(concatpath);
4111
4112        if (addremove != '+')
4113                fill_filespec(one, sha1, mode);
4114        if (addremove != '-') {
4115                fill_filespec(two, sha1, mode);
4116                two->dirty_submodule = dirty_submodule;
4117        }
4118
4119        diff_queue(&diff_queued_diff, one, two);
4120        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4121                DIFF_OPT_SET(options, HAS_CHANGES);
4122}
4123
4124void diff_change(struct diff_options *options,
4125                 unsigned old_mode, unsigned new_mode,
4126                 const unsigned char *old_sha1,
4127                 const unsigned char *new_sha1,
4128                 const char *concatpath,
4129                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4130{
4131        struct diff_filespec *one, *two;
4132
4133        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
4134                        && S_ISGITLINK(new_mode))
4135                return;
4136
4137        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4138                unsigned tmp;
4139                const unsigned char *tmp_c;
4140                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4141                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4142                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4143                        new_dirty_submodule = tmp;
4144        }
4145
4146        if (options->prefix &&
4147            strncmp(concatpath, options->prefix, options->prefix_length))
4148                return;
4149
4150        one = alloc_filespec(concatpath);
4151        two = alloc_filespec(concatpath);
4152        fill_filespec(one, old_sha1, old_mode);
4153        fill_filespec(two, new_sha1, new_mode);
4154        one->dirty_submodule = old_dirty_submodule;
4155        two->dirty_submodule = new_dirty_submodule;
4156
4157        diff_queue(&diff_queued_diff, one, two);
4158        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4159                DIFF_OPT_SET(options, HAS_CHANGES);
4160}
4161
4162void diff_unmerge(struct diff_options *options,
4163                  const char *path,
4164                  unsigned mode, const unsigned char *sha1)
4165{
4166        struct diff_filespec *one, *two;
4167
4168        if (options->prefix &&
4169            strncmp(path, options->prefix, options->prefix_length))
4170                return;
4171
4172        one = alloc_filespec(path);
4173        two = alloc_filespec(path);
4174        fill_filespec(one, sha1, mode);
4175        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
4176}
4177
4178static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4179                size_t *outsize)
4180{
4181        struct diff_tempfile *temp;
4182        const char *argv[3];
4183        const char **arg = argv;
4184        struct child_process child;
4185        struct strbuf buf = STRBUF_INIT;
4186        int err = 0;
4187
4188        temp = prepare_temp_file(spec->path, spec);
4189        *arg++ = pgm;
4190        *arg++ = temp->name;
4191        *arg = NULL;
4192
4193        memset(&child, 0, sizeof(child));
4194        child.use_shell = 1;
4195        child.argv = argv;
4196        child.out = -1;
4197        if (start_command(&child)) {
4198                remove_tempfile();
4199                return NULL;
4200        }
4201
4202        if (strbuf_read(&buf, child.out, 0) < 0)
4203                err = error("error reading from textconv command '%s'", pgm);
4204        close(child.out);
4205
4206        if (finish_command(&child) || err) {
4207                strbuf_release(&buf);
4208                remove_tempfile();
4209                return NULL;
4210        }
4211        remove_tempfile();
4212
4213        return strbuf_detach(&buf, outsize);
4214}
4215
4216static size_t fill_textconv(struct userdiff_driver *driver,
4217                            struct diff_filespec *df,
4218                            char **outbuf)
4219{
4220        size_t size;
4221
4222        if (!driver || !driver->textconv) {
4223                if (!DIFF_FILE_VALID(df)) {
4224                        *outbuf = "";
4225                        return 0;
4226                }
4227                if (diff_populate_filespec(df, 0))
4228                        die("unable to read files to diff");
4229                *outbuf = df->data;
4230                return df->size;
4231        }
4232
4233        if (driver->textconv_cache) {
4234                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4235                                          &size);
4236                if (*outbuf)
4237                        return size;
4238        }
4239
4240        *outbuf = run_textconv(driver->textconv, df, &size);
4241        if (!*outbuf)
4242                die("unable to read files to diff");
4243
4244        if (driver->textconv_cache) {
4245                /* ignore errors, as we might be in a readonly repository */
4246                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4247                                size);
4248                /*
4249                 * we could save up changes and flush them all at the end,
4250                 * but we would need an extra call after all diffing is done.
4251                 * Since generating a cache entry is the slow path anyway,
4252                 * this extra overhead probably isn't a big deal.
4253                 */
4254                notes_cache_write(driver->textconv_cache);
4255        }
4256
4257        return size;
4258}