4a908b28af90e02e4eca8319c1ac51bb788a27c1
   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
  14#ifdef NO_FAST_WORKING_DIRECTORY
  15#define FAST_WORKING_DIRECTORY 0
  16#else
  17#define FAST_WORKING_DIRECTORY 1
  18#endif
  19
  20static int diff_detect_rename_default;
  21static int diff_rename_limit_default = 100;
  22static int diff_use_color_default;
  23static const char *external_diff_cmd_cfg;
  24int diff_auto_refresh_index = 1;
  25
  26static char diff_colors[][COLOR_MAXLEN] = {
  27        "\033[m",       /* reset */
  28        "",             /* PLAIN (normal) */
  29        "\033[1m",      /* METAINFO (bold) */
  30        "\033[36m",     /* FRAGINFO (cyan) */
  31        "\033[31m",     /* OLD (red) */
  32        "\033[32m",     /* NEW (green) */
  33        "\033[33m",     /* COMMIT (yellow) */
  34        "\033[41m",     /* WHITESPACE (red background) */
  35};
  36
  37static int parse_diff_color_slot(const char *var, int ofs)
  38{
  39        if (!strcasecmp(var+ofs, "plain"))
  40                return DIFF_PLAIN;
  41        if (!strcasecmp(var+ofs, "meta"))
  42                return DIFF_METAINFO;
  43        if (!strcasecmp(var+ofs, "frag"))
  44                return DIFF_FRAGINFO;
  45        if (!strcasecmp(var+ofs, "old"))
  46                return DIFF_FILE_OLD;
  47        if (!strcasecmp(var+ofs, "new"))
  48                return DIFF_FILE_NEW;
  49        if (!strcasecmp(var+ofs, "commit"))
  50                return DIFF_COMMIT;
  51        if (!strcasecmp(var+ofs, "whitespace"))
  52                return DIFF_WHITESPACE;
  53        die("bad config variable '%s'", var);
  54}
  55
  56static struct ll_diff_driver {
  57        const char *name;
  58        struct ll_diff_driver *next;
  59        char *cmd;
  60} *user_diff, **user_diff_tail;
  61
  62static void read_config_if_needed(void)
  63{
  64        if (!user_diff_tail) {
  65                user_diff_tail = &user_diff;
  66                git_config(git_diff_ui_config);
  67        }
  68}
  69
  70/*
  71 * Currently there is only "diff.<drivername>.command" variable;
  72 * because there are "diff.color.<slot>" variables, we are parsing
  73 * this in a bit convoluted way to allow low level diff driver
  74 * called "color".
  75 */
  76static int parse_lldiff_command(const char *var, const char *ep, const char *value)
  77{
  78        const char *name;
  79        int namelen;
  80        struct ll_diff_driver *drv;
  81
  82        name = var + 5;
  83        namelen = ep - name;
  84        for (drv = user_diff; drv; drv = drv->next)
  85                if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
  86                        break;
  87        if (!drv) {
  88                drv = xcalloc(1, sizeof(struct ll_diff_driver));
  89                drv->name = xmemdupz(name, namelen);
  90                if (!user_diff_tail)
  91                        user_diff_tail = &user_diff;
  92                *user_diff_tail = drv;
  93                user_diff_tail = &(drv->next);
  94        }
  95
  96        if (!value)
  97                return error("%s: lacks value", var);
  98        drv->cmd = strdup(value);
  99        return 0;
 100}
 101
 102/*
 103 * 'diff.<what>.funcname' attribute can be specified in the configuration
 104 * to define a customized regexp to find the beginning of a function to
 105 * be used for hunk header lines of "diff -p" style output.
 106 */
 107static struct funcname_pattern {
 108        char *name;
 109        char *pattern;
 110        struct funcname_pattern *next;
 111} *funcname_pattern_list;
 112
 113static int parse_funcname_pattern(const char *var, const char *ep, const char *value)
 114{
 115        const char *name;
 116        int namelen;
 117        struct funcname_pattern *pp;
 118
 119        name = var + 5; /* "diff." */
 120        namelen = ep - name;
 121
 122        for (pp = funcname_pattern_list; pp; pp = pp->next)
 123                if (!strncmp(pp->name, name, namelen) && !pp->name[namelen])
 124                        break;
 125        if (!pp) {
 126                pp = xcalloc(1, sizeof(*pp));
 127                pp->name = xmemdupz(name, namelen);
 128                pp->next = funcname_pattern_list;
 129                funcname_pattern_list = pp;
 130        }
 131        if (pp->pattern)
 132                free(pp->pattern);
 133        pp->pattern = xstrdup(value);
 134        return 0;
 135}
 136
 137/*
 138 * These are to give UI layer defaults.
 139 * The core-level commands such as git-diff-files should
 140 * never be affected by the setting of diff.renames
 141 * the user happens to have in the configuration file.
 142 */
 143int git_diff_ui_config(const char *var, const char *value)
 144{
 145        if (!strcmp(var, "diff.renamelimit")) {
 146                diff_rename_limit_default = git_config_int(var, value);
 147                return 0;
 148        }
 149        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 150                diff_use_color_default = git_config_colorbool(var, value, -1);
 151                return 0;
 152        }
 153        if (!strcmp(var, "diff.renames")) {
 154                if (!value)
 155                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 156                else if (!strcasecmp(value, "copies") ||
 157                         !strcasecmp(value, "copy"))
 158                        diff_detect_rename_default = DIFF_DETECT_COPY;
 159                else if (git_config_bool(var,value))
 160                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 161                return 0;
 162        }
 163        if (!strcmp(var, "diff.autorefreshindex")) {
 164                diff_auto_refresh_index = git_config_bool(var, value);
 165                return 0;
 166        }
 167        if (!strcmp(var, "diff.external")) {
 168                external_diff_cmd_cfg = xstrdup(value);
 169                return 0;
 170        }
 171        if (!prefixcmp(var, "diff.")) {
 172                const char *ep = strrchr(var, '.');
 173
 174                if (ep != var + 4) {
 175                        if (!strcmp(ep, ".command"))
 176                                return parse_lldiff_command(var, ep, value);
 177                }
 178        }
 179
 180        return git_diff_basic_config(var, value);
 181}
 182
 183int git_diff_basic_config(const char *var, const char *value)
 184{
 185        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 186                int slot = parse_diff_color_slot(var, 11);
 187                color_parse(value, var, diff_colors[slot]);
 188                return 0;
 189        }
 190
 191        if (!prefixcmp(var, "diff.")) {
 192                const char *ep = strrchr(var, '.');
 193                if (ep != var + 4) {
 194                        if (!strcmp(ep, ".funcname"))
 195                                return parse_funcname_pattern(var, ep, value);
 196                }
 197        }
 198
 199        return git_default_config(var, value);
 200}
 201
 202static char *quote_two(const char *one, const char *two)
 203{
 204        int need_one = quote_c_style(one, NULL, NULL, 1);
 205        int need_two = quote_c_style(two, NULL, NULL, 1);
 206        struct strbuf res;
 207
 208        strbuf_init(&res, 0);
 209        if (need_one + need_two) {
 210                strbuf_addch(&res, '"');
 211                quote_c_style(one, &res, NULL, 1);
 212                quote_c_style(two, &res, NULL, 1);
 213                strbuf_addch(&res, '"');
 214        } else {
 215                strbuf_addstr(&res, one);
 216                strbuf_addstr(&res, two);
 217        }
 218        return strbuf_detach(&res, NULL);
 219}
 220
 221static const char *external_diff(void)
 222{
 223        static const char *external_diff_cmd = NULL;
 224        static int done_preparing = 0;
 225
 226        if (done_preparing)
 227                return external_diff_cmd;
 228        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 229        if (!external_diff_cmd)
 230                external_diff_cmd = external_diff_cmd_cfg;
 231        done_preparing = 1;
 232        return external_diff_cmd;
 233}
 234
 235static struct diff_tempfile {
 236        const char *name; /* filename external diff should read from */
 237        char hex[41];
 238        char mode[10];
 239        char tmp_path[PATH_MAX];
 240} diff_temp[2];
 241
 242static int count_lines(const char *data, int size)
 243{
 244        int count, ch, completely_empty = 1, nl_just_seen = 0;
 245        count = 0;
 246        while (0 < size--) {
 247                ch = *data++;
 248                if (ch == '\n') {
 249                        count++;
 250                        nl_just_seen = 1;
 251                        completely_empty = 0;
 252                }
 253                else {
 254                        nl_just_seen = 0;
 255                        completely_empty = 0;
 256                }
 257        }
 258        if (completely_empty)
 259                return 0;
 260        if (!nl_just_seen)
 261                count++; /* no trailing newline */
 262        return count;
 263}
 264
 265static void print_line_count(int count)
 266{
 267        switch (count) {
 268        case 0:
 269                printf("0,0");
 270                break;
 271        case 1:
 272                printf("1");
 273                break;
 274        default:
 275                printf("1,%d", count);
 276                break;
 277        }
 278}
 279
 280static void copy_file(int prefix, const char *data, int size,
 281                const char *set, const char *reset)
 282{
 283        int ch, nl_just_seen = 1;
 284        while (0 < size--) {
 285                ch = *data++;
 286                if (nl_just_seen) {
 287                        fputs(set, stdout);
 288                        putchar(prefix);
 289                }
 290                if (ch == '\n') {
 291                        nl_just_seen = 1;
 292                        fputs(reset, stdout);
 293                } else
 294                        nl_just_seen = 0;
 295                putchar(ch);
 296        }
 297        if (!nl_just_seen)
 298                printf("%s\n\\ No newline at end of file\n", reset);
 299}
 300
 301static void emit_rewrite_diff(const char *name_a,
 302                              const char *name_b,
 303                              struct diff_filespec *one,
 304                              struct diff_filespec *two,
 305                              struct diff_options *o)
 306{
 307        int lc_a, lc_b;
 308        int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
 309        const char *name_a_tab, *name_b_tab;
 310        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 311        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 312        const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
 313        const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
 314        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 315        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 316
 317        name_a += (*name_a == '/');
 318        name_b += (*name_b == '/');
 319        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 320        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 321
 322        strbuf_reset(&a_name);
 323        strbuf_reset(&b_name);
 324        quote_two_c_style(&a_name, o->a_prefix, name_a, 0);
 325        quote_two_c_style(&b_name, o->b_prefix, name_b, 0);
 326
 327        diff_populate_filespec(one, 0);
 328        diff_populate_filespec(two, 0);
 329        lc_a = count_lines(one->data, one->size);
 330        lc_b = count_lines(two->data, two->size);
 331        printf("%s--- %s%s%s\n%s+++ %s%s%s\n%s@@ -",
 332               metainfo, a_name.buf, name_a_tab, reset,
 333               metainfo, b_name.buf, name_b_tab, reset, fraginfo);
 334        print_line_count(lc_a);
 335        printf(" +");
 336        print_line_count(lc_b);
 337        printf(" @@%s\n", reset);
 338        if (lc_a)
 339                copy_file('-', one->data, one->size, old, reset);
 340        if (lc_b)
 341                copy_file('+', two->data, two->size, new, reset);
 342}
 343
 344static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 345{
 346        if (!DIFF_FILE_VALID(one)) {
 347                mf->ptr = (char *)""; /* does not matter */
 348                mf->size = 0;
 349                return 0;
 350        }
 351        else if (diff_populate_filespec(one, 0))
 352                return -1;
 353        mf->ptr = one->data;
 354        mf->size = one->size;
 355        return 0;
 356}
 357
 358struct diff_words_buffer {
 359        mmfile_t text;
 360        long alloc;
 361        long current; /* output pointer */
 362        int suppressed_newline;
 363};
 364
 365static void diff_words_append(char *line, unsigned long len,
 366                struct diff_words_buffer *buffer)
 367{
 368        if (buffer->text.size + len > buffer->alloc) {
 369                buffer->alloc = (buffer->text.size + len) * 3 / 2;
 370                buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
 371        }
 372        line++;
 373        len--;
 374        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 375        buffer->text.size += len;
 376}
 377
 378struct diff_words_data {
 379        struct xdiff_emit_state xm;
 380        struct diff_words_buffer minus, plus;
 381};
 382
 383static void print_word(struct diff_words_buffer *buffer, int len, int color,
 384                int suppress_newline)
 385{
 386        const char *ptr;
 387        int eol = 0;
 388
 389        if (len == 0)
 390                return;
 391
 392        ptr  = buffer->text.ptr + buffer->current;
 393        buffer->current += len;
 394
 395        if (ptr[len - 1] == '\n') {
 396                eol = 1;
 397                len--;
 398        }
 399
 400        fputs(diff_get_color(1, color), stdout);
 401        fwrite(ptr, len, 1, stdout);
 402        fputs(diff_get_color(1, DIFF_RESET), stdout);
 403
 404        if (eol) {
 405                if (suppress_newline)
 406                        buffer->suppressed_newline = 1;
 407                else
 408                        putchar('\n');
 409        }
 410}
 411
 412static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 413{
 414        struct diff_words_data *diff_words = priv;
 415
 416        if (diff_words->minus.suppressed_newline) {
 417                if (line[0] != '+')
 418                        putchar('\n');
 419                diff_words->minus.suppressed_newline = 0;
 420        }
 421
 422        len--;
 423        switch (line[0]) {
 424                case '-':
 425                        print_word(&diff_words->minus, len, DIFF_FILE_OLD, 1);
 426                        break;
 427                case '+':
 428                        print_word(&diff_words->plus, len, DIFF_FILE_NEW, 0);
 429                        break;
 430                case ' ':
 431                        print_word(&diff_words->plus, len, DIFF_PLAIN, 0);
 432                        diff_words->minus.current += len;
 433                        break;
 434        }
 435}
 436
 437/* this executes the word diff on the accumulated buffers */
 438static void diff_words_show(struct diff_words_data *diff_words)
 439{
 440        xpparam_t xpp;
 441        xdemitconf_t xecfg;
 442        xdemitcb_t ecb;
 443        mmfile_t minus, plus;
 444        int i;
 445
 446        memset(&xecfg, 0, sizeof(xecfg));
 447        minus.size = diff_words->minus.text.size;
 448        minus.ptr = xmalloc(minus.size);
 449        memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
 450        for (i = 0; i < minus.size; i++)
 451                if (isspace(minus.ptr[i]))
 452                        minus.ptr[i] = '\n';
 453        diff_words->minus.current = 0;
 454
 455        plus.size = diff_words->plus.text.size;
 456        plus.ptr = xmalloc(plus.size);
 457        memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
 458        for (i = 0; i < plus.size; i++)
 459                if (isspace(plus.ptr[i]))
 460                        plus.ptr[i] = '\n';
 461        diff_words->plus.current = 0;
 462
 463        xpp.flags = XDF_NEED_MINIMAL;
 464        xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
 465        ecb.outf = xdiff_outf;
 466        ecb.priv = diff_words;
 467        diff_words->xm.consume = fn_out_diff_words_aux;
 468        xdi_diff(&minus, &plus, &xpp, &xecfg, &ecb);
 469
 470        free(minus.ptr);
 471        free(plus.ptr);
 472        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 473
 474        if (diff_words->minus.suppressed_newline) {
 475                putchar('\n');
 476                diff_words->minus.suppressed_newline = 0;
 477        }
 478}
 479
 480struct emit_callback {
 481        struct xdiff_emit_state xm;
 482        int nparents, color_diff;
 483        unsigned ws_rule;
 484        const char **label_path;
 485        struct diff_words_data *diff_words;
 486        int *found_changesp;
 487};
 488
 489static void free_diff_words_data(struct emit_callback *ecbdata)
 490{
 491        if (ecbdata->diff_words) {
 492                /* flush buffers */
 493                if (ecbdata->diff_words->minus.text.size ||
 494                                ecbdata->diff_words->plus.text.size)
 495                        diff_words_show(ecbdata->diff_words);
 496
 497                if (ecbdata->diff_words->minus.text.ptr)
 498                        free (ecbdata->diff_words->minus.text.ptr);
 499                if (ecbdata->diff_words->plus.text.ptr)
 500                        free (ecbdata->diff_words->plus.text.ptr);
 501                free(ecbdata->diff_words);
 502                ecbdata->diff_words = NULL;
 503        }
 504}
 505
 506const char *diff_get_color(int diff_use_color, enum color_diff ix)
 507{
 508        if (diff_use_color)
 509                return diff_colors[ix];
 510        return "";
 511}
 512
 513static void emit_line(const char *set, const char *reset, const char *line, int len)
 514{
 515        fputs(set, stdout);
 516        fwrite(line, len, 1, stdout);
 517        fputs(reset, stdout);
 518}
 519
 520static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 521{
 522        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 523        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 524
 525        if (!*ws)
 526                emit_line(set, reset, line, len);
 527        else {
 528                /* Emit just the prefix, then the rest. */
 529                emit_line(set, reset, line, ecbdata->nparents);
 530                (void)check_and_emit_line(line + ecbdata->nparents,
 531                    len - ecbdata->nparents, ecbdata->ws_rule,
 532                    stdout, set, reset, ws);
 533        }
 534}
 535
 536static void fn_out_consume(void *priv, char *line, unsigned long len)
 537{
 538        int i;
 539        int color;
 540        struct emit_callback *ecbdata = priv;
 541        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 542        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 543
 544        *(ecbdata->found_changesp) = 1;
 545
 546        if (ecbdata->label_path[0]) {
 547                const char *name_a_tab, *name_b_tab;
 548
 549                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 550                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 551
 552                printf("%s--- %s%s%s\n",
 553                       set, ecbdata->label_path[0], reset, name_a_tab);
 554                printf("%s+++ %s%s%s\n",
 555                       set, ecbdata->label_path[1], reset, name_b_tab);
 556                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 557        }
 558
 559        /* This is not really necessary for now because
 560         * this codepath only deals with two-way diffs.
 561         */
 562        for (i = 0; i < len && line[i] == '@'; i++)
 563                ;
 564        if (2 <= i && i < len && line[i] == ' ') {
 565                ecbdata->nparents = i - 1;
 566                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 567                          reset, line, len);
 568                return;
 569        }
 570
 571        if (len < ecbdata->nparents) {
 572                set = reset;
 573                emit_line(reset, reset, line, len);
 574                return;
 575        }
 576
 577        color = DIFF_PLAIN;
 578        if (ecbdata->diff_words && ecbdata->nparents != 1)
 579                /* fall back to normal diff */
 580                free_diff_words_data(ecbdata);
 581        if (ecbdata->diff_words) {
 582                if (line[0] == '-') {
 583                        diff_words_append(line, len,
 584                                          &ecbdata->diff_words->minus);
 585                        return;
 586                } else if (line[0] == '+') {
 587                        diff_words_append(line, len,
 588                                          &ecbdata->diff_words->plus);
 589                        return;
 590                }
 591                if (ecbdata->diff_words->minus.text.size ||
 592                    ecbdata->diff_words->plus.text.size)
 593                        diff_words_show(ecbdata->diff_words);
 594                line++;
 595                len--;
 596                emit_line(set, reset, line, len);
 597                return;
 598        }
 599        for (i = 0; i < ecbdata->nparents && len; i++) {
 600                if (line[i] == '-')
 601                        color = DIFF_FILE_OLD;
 602                else if (line[i] == '+')
 603                        color = DIFF_FILE_NEW;
 604        }
 605
 606        if (color != DIFF_FILE_NEW) {
 607                emit_line(diff_get_color(ecbdata->color_diff, color),
 608                          reset, line, len);
 609                return;
 610        }
 611        emit_add_line(reset, ecbdata, line, len);
 612}
 613
 614static char *pprint_rename(const char *a, const char *b)
 615{
 616        const char *old = a;
 617        const char *new = b;
 618        struct strbuf name;
 619        int pfx_length, sfx_length;
 620        int len_a = strlen(a);
 621        int len_b = strlen(b);
 622        int a_midlen, b_midlen;
 623        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 624        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 625
 626        strbuf_init(&name, 0);
 627        if (qlen_a || qlen_b) {
 628                quote_c_style(a, &name, NULL, 0);
 629                strbuf_addstr(&name, " => ");
 630                quote_c_style(b, &name, NULL, 0);
 631                return strbuf_detach(&name, NULL);
 632        }
 633
 634        /* Find common prefix */
 635        pfx_length = 0;
 636        while (*old && *new && *old == *new) {
 637                if (*old == '/')
 638                        pfx_length = old - a + 1;
 639                old++;
 640                new++;
 641        }
 642
 643        /* Find common suffix */
 644        old = a + len_a;
 645        new = b + len_b;
 646        sfx_length = 0;
 647        while (a <= old && b <= new && *old == *new) {
 648                if (*old == '/')
 649                        sfx_length = len_a - (old - a);
 650                old--;
 651                new--;
 652        }
 653
 654        /*
 655         * pfx{mid-a => mid-b}sfx
 656         * {pfx-a => pfx-b}sfx
 657         * pfx{sfx-a => sfx-b}
 658         * name-a => name-b
 659         */
 660        a_midlen = len_a - pfx_length - sfx_length;
 661        b_midlen = len_b - pfx_length - sfx_length;
 662        if (a_midlen < 0)
 663                a_midlen = 0;
 664        if (b_midlen < 0)
 665                b_midlen = 0;
 666
 667        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 668        if (pfx_length + sfx_length) {
 669                strbuf_add(&name, a, pfx_length);
 670                strbuf_addch(&name, '{');
 671        }
 672        strbuf_add(&name, a + pfx_length, a_midlen);
 673        strbuf_addstr(&name, " => ");
 674        strbuf_add(&name, b + pfx_length, b_midlen);
 675        if (pfx_length + sfx_length) {
 676                strbuf_addch(&name, '}');
 677                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 678        }
 679        return strbuf_detach(&name, NULL);
 680}
 681
 682struct diffstat_t {
 683        struct xdiff_emit_state xm;
 684
 685        int nr;
 686        int alloc;
 687        struct diffstat_file {
 688                char *from_name;
 689                char *name;
 690                char *print_name;
 691                unsigned is_unmerged:1;
 692                unsigned is_binary:1;
 693                unsigned is_renamed:1;
 694                unsigned int added, deleted;
 695        } **files;
 696};
 697
 698static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 699                                          const char *name_a,
 700                                          const char *name_b)
 701{
 702        struct diffstat_file *x;
 703        x = xcalloc(sizeof (*x), 1);
 704        if (diffstat->nr == diffstat->alloc) {
 705                diffstat->alloc = alloc_nr(diffstat->alloc);
 706                diffstat->files = xrealloc(diffstat->files,
 707                                diffstat->alloc * sizeof(x));
 708        }
 709        diffstat->files[diffstat->nr++] = x;
 710        if (name_b) {
 711                x->from_name = xstrdup(name_a);
 712                x->name = xstrdup(name_b);
 713                x->is_renamed = 1;
 714        }
 715        else {
 716                x->from_name = NULL;
 717                x->name = xstrdup(name_a);
 718        }
 719        return x;
 720}
 721
 722static void diffstat_consume(void *priv, char *line, unsigned long len)
 723{
 724        struct diffstat_t *diffstat = priv;
 725        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 726
 727        if (line[0] == '+')
 728                x->added++;
 729        else if (line[0] == '-')
 730                x->deleted++;
 731}
 732
 733const char mime_boundary_leader[] = "------------";
 734
 735static int scale_linear(int it, int width, int max_change)
 736{
 737        /*
 738         * make sure that at least one '-' is printed if there were deletions,
 739         * and likewise for '+'.
 740         */
 741        if (max_change < 2)
 742                return it;
 743        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 744}
 745
 746static void show_name(const char *prefix, const char *name, int len,
 747                      const char *reset, const char *set)
 748{
 749        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 750}
 751
 752static void show_graph(char ch, int cnt, const char *set, const char *reset)
 753{
 754        if (cnt <= 0)
 755                return;
 756        printf("%s", set);
 757        while (cnt--)
 758                putchar(ch);
 759        printf("%s", reset);
 760}
 761
 762static void fill_print_name(struct diffstat_file *file)
 763{
 764        char *pname;
 765
 766        if (file->print_name)
 767                return;
 768
 769        if (!file->is_renamed) {
 770                struct strbuf buf;
 771                strbuf_init(&buf, 0);
 772                if (quote_c_style(file->name, &buf, NULL, 0)) {
 773                        pname = strbuf_detach(&buf, NULL);
 774                } else {
 775                        pname = file->name;
 776                        strbuf_release(&buf);
 777                }
 778        } else {
 779                pname = pprint_rename(file->from_name, file->name);
 780        }
 781        file->print_name = pname;
 782}
 783
 784static void show_stats(struct diffstat_t* data, struct diff_options *options)
 785{
 786        int i, len, add, del, total, adds = 0, dels = 0;
 787        int max_change = 0, max_len = 0;
 788        int total_files = data->nr;
 789        int width, name_width;
 790        const char *reset, *set, *add_c, *del_c;
 791
 792        if (data->nr == 0)
 793                return;
 794
 795        width = options->stat_width ? options->stat_width : 80;
 796        name_width = options->stat_name_width ? options->stat_name_width : 50;
 797
 798        /* Sanity: give at least 5 columns to the graph,
 799         * but leave at least 10 columns for the name.
 800         */
 801        if (width < name_width + 15) {
 802                if (name_width <= 25)
 803                        width = name_width + 15;
 804                else
 805                        name_width = width - 15;
 806        }
 807
 808        /* Find the longest filename and max number of changes */
 809        reset = diff_get_color_opt(options, DIFF_RESET);
 810        set   = diff_get_color_opt(options, DIFF_PLAIN);
 811        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
 812        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
 813
 814        for (i = 0; i < data->nr; i++) {
 815                struct diffstat_file *file = data->files[i];
 816                int change = file->added + file->deleted;
 817                fill_print_name(file);
 818                len = strlen(file->print_name);
 819                if (max_len < len)
 820                        max_len = len;
 821
 822                if (file->is_binary || file->is_unmerged)
 823                        continue;
 824                if (max_change < change)
 825                        max_change = change;
 826        }
 827
 828        /* Compute the width of the graph part;
 829         * 10 is for one blank at the beginning of the line plus
 830         * " | count " between the name and the graph.
 831         *
 832         * From here on, name_width is the width of the name area,
 833         * and width is the width of the graph area.
 834         */
 835        name_width = (name_width < max_len) ? name_width : max_len;
 836        if (width < (name_width + 10) + max_change)
 837                width = width - (name_width + 10);
 838        else
 839                width = max_change;
 840
 841        for (i = 0; i < data->nr; i++) {
 842                const char *prefix = "";
 843                char *name = data->files[i]->print_name;
 844                int added = data->files[i]->added;
 845                int deleted = data->files[i]->deleted;
 846                int name_len;
 847
 848                /*
 849                 * "scale" the filename
 850                 */
 851                len = name_width;
 852                name_len = strlen(name);
 853                if (name_width < name_len) {
 854                        char *slash;
 855                        prefix = "...";
 856                        len -= 3;
 857                        name += name_len - len;
 858                        slash = strchr(name, '/');
 859                        if (slash)
 860                                name = slash;
 861                }
 862
 863                if (data->files[i]->is_binary) {
 864                        show_name(prefix, name, len, reset, set);
 865                        printf("  Bin ");
 866                        printf("%s%d%s", del_c, deleted, reset);
 867                        printf(" -> ");
 868                        printf("%s%d%s", add_c, added, reset);
 869                        printf(" bytes");
 870                        printf("\n");
 871                        continue;
 872                }
 873                else if (data->files[i]->is_unmerged) {
 874                        show_name(prefix, name, len, reset, set);
 875                        printf("  Unmerged\n");
 876                        continue;
 877                }
 878                else if (!data->files[i]->is_renamed &&
 879                         (added + deleted == 0)) {
 880                        total_files--;
 881                        continue;
 882                }
 883
 884                /*
 885                 * scale the add/delete
 886                 */
 887                add = added;
 888                del = deleted;
 889                total = add + del;
 890                adds += add;
 891                dels += del;
 892
 893                if (width <= max_change) {
 894                        add = scale_linear(add, width, max_change);
 895                        del = scale_linear(del, width, max_change);
 896                        total = add + del;
 897                }
 898                show_name(prefix, name, len, reset, set);
 899                printf("%5d ", added + deleted);
 900                show_graph('+', add, add_c, reset);
 901                show_graph('-', del, del_c, reset);
 902                putchar('\n');
 903        }
 904        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 905               set, total_files, adds, dels, reset);
 906}
 907
 908static void show_shortstats(struct diffstat_t* data)
 909{
 910        int i, adds = 0, dels = 0, total_files = data->nr;
 911
 912        if (data->nr == 0)
 913                return;
 914
 915        for (i = 0; i < data->nr; i++) {
 916                if (!data->files[i]->is_binary &&
 917                    !data->files[i]->is_unmerged) {
 918                        int added = data->files[i]->added;
 919                        int deleted= data->files[i]->deleted;
 920                        if (!data->files[i]->is_renamed &&
 921                            (added + deleted == 0)) {
 922                                total_files--;
 923                        } else {
 924                                adds += added;
 925                                dels += deleted;
 926                        }
 927                }
 928        }
 929        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 930               total_files, adds, dels);
 931}
 932
 933static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 934{
 935        int i;
 936
 937        if (data->nr == 0)
 938                return;
 939
 940        for (i = 0; i < data->nr; i++) {
 941                struct diffstat_file *file = data->files[i];
 942
 943                if (file->is_binary)
 944                        printf("-\t-\t");
 945                else
 946                        printf("%d\t%d\t", file->added, file->deleted);
 947                if (options->line_termination) {
 948                        fill_print_name(file);
 949                        if (!file->is_renamed)
 950                                write_name_quoted(file->name, stdout,
 951                                                  options->line_termination);
 952                        else {
 953                                fputs(file->print_name, stdout);
 954                                putchar(options->line_termination);
 955                        }
 956                } else {
 957                        if (file->is_renamed) {
 958                                putchar('\0');
 959                                write_name_quoted(file->from_name, stdout, '\0');
 960                        }
 961                        write_name_quoted(file->name, stdout, '\0');
 962                }
 963        }
 964}
 965
 966static void free_diffstat_info(struct diffstat_t *diffstat)
 967{
 968        int i;
 969        for (i = 0; i < diffstat->nr; i++) {
 970                struct diffstat_file *f = diffstat->files[i];
 971                if (f->name != f->print_name)
 972                        free(f->print_name);
 973                free(f->name);
 974                free(f->from_name);
 975                free(f);
 976        }
 977        free(diffstat->files);
 978}
 979
 980struct checkdiff_t {
 981        struct xdiff_emit_state xm;
 982        const char *filename;
 983        int lineno, color_diff;
 984        unsigned ws_rule;
 985        unsigned status;
 986};
 987
 988static void checkdiff_consume(void *priv, char *line, unsigned long len)
 989{
 990        struct checkdiff_t *data = priv;
 991        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
 992        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
 993        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
 994        char *err;
 995
 996        if (line[0] == '+') {
 997                data->status = check_and_emit_line(line + 1, len - 1,
 998                    data->ws_rule, NULL, NULL, NULL, NULL);
 999                if (!data->status)
1000                        return;
1001                err = whitespace_error_string(data->status);
1002                printf("%s:%d: %s.\n", data->filename, data->lineno, err);
1003                free(err);
1004                emit_line(set, reset, line, 1);
1005                (void)check_and_emit_line(line + 1, len - 1, data->ws_rule,
1006                    stdout, set, reset, ws);
1007                data->lineno++;
1008        } else if (line[0] == ' ')
1009                data->lineno++;
1010        else if (line[0] == '@') {
1011                char *plus = strchr(line, '+');
1012                if (plus)
1013                        data->lineno = strtol(plus, NULL, 10);
1014                else
1015                        die("invalid diff");
1016        }
1017}
1018
1019static unsigned char *deflate_it(char *data,
1020                                 unsigned long size,
1021                                 unsigned long *result_size)
1022{
1023        int bound;
1024        unsigned char *deflated;
1025        z_stream stream;
1026
1027        memset(&stream, 0, sizeof(stream));
1028        deflateInit(&stream, zlib_compression_level);
1029        bound = deflateBound(&stream, size);
1030        deflated = xmalloc(bound);
1031        stream.next_out = deflated;
1032        stream.avail_out = bound;
1033
1034        stream.next_in = (unsigned char *)data;
1035        stream.avail_in = size;
1036        while (deflate(&stream, Z_FINISH) == Z_OK)
1037                ; /* nothing */
1038        deflateEnd(&stream);
1039        *result_size = stream.total_out;
1040        return deflated;
1041}
1042
1043static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1044{
1045        void *cp;
1046        void *delta;
1047        void *deflated;
1048        void *data;
1049        unsigned long orig_size;
1050        unsigned long delta_size;
1051        unsigned long deflate_size;
1052        unsigned long data_size;
1053
1054        /* We could do deflated delta, or we could do just deflated two,
1055         * whichever is smaller.
1056         */
1057        delta = NULL;
1058        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1059        if (one->size && two->size) {
1060                delta = diff_delta(one->ptr, one->size,
1061                                   two->ptr, two->size,
1062                                   &delta_size, deflate_size);
1063                if (delta) {
1064                        void *to_free = delta;
1065                        orig_size = delta_size;
1066                        delta = deflate_it(delta, delta_size, &delta_size);
1067                        free(to_free);
1068                }
1069        }
1070
1071        if (delta && delta_size < deflate_size) {
1072                printf("delta %lu\n", orig_size);
1073                free(deflated);
1074                data = delta;
1075                data_size = delta_size;
1076        }
1077        else {
1078                printf("literal %lu\n", two->size);
1079                free(delta);
1080                data = deflated;
1081                data_size = deflate_size;
1082        }
1083
1084        /* emit data encoded in base85 */
1085        cp = data;
1086        while (data_size) {
1087                int bytes = (52 < data_size) ? 52 : data_size;
1088                char line[70];
1089                data_size -= bytes;
1090                if (bytes <= 26)
1091                        line[0] = bytes + 'A' - 1;
1092                else
1093                        line[0] = bytes - 26 + 'a' - 1;
1094                encode_85(line + 1, cp, bytes);
1095                cp = (char *) cp + bytes;
1096                puts(line);
1097        }
1098        printf("\n");
1099        free(data);
1100}
1101
1102static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1103{
1104        printf("GIT binary patch\n");
1105        emit_binary_diff_body(one, two);
1106        emit_binary_diff_body(two, one);
1107}
1108
1109static void setup_diff_attr_check(struct git_attr_check *check)
1110{
1111        static struct git_attr *attr_diff;
1112
1113        if (!attr_diff) {
1114                attr_diff = git_attr("diff", 4);
1115        }
1116        check[0].attr = attr_diff;
1117}
1118
1119static void diff_filespec_check_attr(struct diff_filespec *one)
1120{
1121        struct git_attr_check attr_diff_check;
1122        int check_from_data = 0;
1123
1124        if (one->checked_attr)
1125                return;
1126
1127        setup_diff_attr_check(&attr_diff_check);
1128        one->is_binary = 0;
1129        one->funcname_pattern_ident = NULL;
1130
1131        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1132                const char *value;
1133
1134                /* binaryness */
1135                value = attr_diff_check.value;
1136                if (ATTR_TRUE(value))
1137                        ;
1138                else if (ATTR_FALSE(value))
1139                        one->is_binary = 1;
1140                else
1141                        check_from_data = 1;
1142
1143                /* funcname pattern ident */
1144                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1145                        ;
1146                else
1147                        one->funcname_pattern_ident = value;
1148        }
1149
1150        if (check_from_data) {
1151                if (!one->data && DIFF_FILE_VALID(one))
1152                        diff_populate_filespec(one, 0);
1153
1154                if (one->data)
1155                        one->is_binary = buffer_is_binary(one->data, one->size);
1156        }
1157}
1158
1159int diff_filespec_is_binary(struct diff_filespec *one)
1160{
1161        diff_filespec_check_attr(one);
1162        return one->is_binary;
1163}
1164
1165static const char *funcname_pattern(const char *ident)
1166{
1167        struct funcname_pattern *pp;
1168
1169        for (pp = funcname_pattern_list; pp; pp = pp->next)
1170                if (!strcmp(ident, pp->name))
1171                        return pp->pattern;
1172        return NULL;
1173}
1174
1175static struct builtin_funcname_pattern {
1176        const char *name;
1177        const char *pattern;
1178} builtin_funcname_pattern[] = {
1179        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1180                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1181                        "^[     ]*\\(\\([       ]*"
1182                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1183                        "[      ]*([^;]*$\\)" },
1184        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1185};
1186
1187static const char *diff_funcname_pattern(struct diff_filespec *one)
1188{
1189        const char *ident, *pattern;
1190        int i;
1191
1192        diff_filespec_check_attr(one);
1193        ident = one->funcname_pattern_ident;
1194
1195        if (!ident)
1196                /*
1197                 * If the config file has "funcname.default" defined, that
1198                 * regexp is used; otherwise NULL is returned and xemit uses
1199                 * the built-in default.
1200                 */
1201                return funcname_pattern("default");
1202
1203        /* Look up custom "funcname.$ident" regexp from config. */
1204        pattern = funcname_pattern(ident);
1205        if (pattern)
1206                return pattern;
1207
1208        /*
1209         * And define built-in fallback patterns here.  Note that
1210         * these can be overriden by the user's config settings.
1211         */
1212        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1213                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1214                        return builtin_funcname_pattern[i].pattern;
1215
1216        return NULL;
1217}
1218
1219static void builtin_diff(const char *name_a,
1220                         const char *name_b,
1221                         struct diff_filespec *one,
1222                         struct diff_filespec *two,
1223                         const char *xfrm_msg,
1224                         struct diff_options *o,
1225                         int complete_rewrite)
1226{
1227        mmfile_t mf1, mf2;
1228        const char *lbl[2];
1229        char *a_one, *b_two;
1230        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1231        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1232
1233        a_one = quote_two(o->a_prefix, name_a + (*name_a == '/'));
1234        b_two = quote_two(o->b_prefix, name_b + (*name_b == '/'));
1235        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1236        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1237        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1238        if (lbl[0][0] == '/') {
1239                /* /dev/null */
1240                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1241                if (xfrm_msg && xfrm_msg[0])
1242                        printf("%s%s%s\n", set, xfrm_msg, reset);
1243        }
1244        else if (lbl[1][0] == '/') {
1245                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1246                if (xfrm_msg && xfrm_msg[0])
1247                        printf("%s%s%s\n", set, xfrm_msg, reset);
1248        }
1249        else {
1250                if (one->mode != two->mode) {
1251                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1252                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1253                }
1254                if (xfrm_msg && xfrm_msg[0])
1255                        printf("%s%s%s\n", set, xfrm_msg, reset);
1256                /*
1257                 * we do not run diff between different kind
1258                 * of objects.
1259                 */
1260                if ((one->mode ^ two->mode) & S_IFMT)
1261                        goto free_ab_and_return;
1262                if (complete_rewrite) {
1263                        emit_rewrite_diff(name_a, name_b, one, two, o);
1264                        o->found_changes = 1;
1265                        goto free_ab_and_return;
1266                }
1267        }
1268
1269        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1270                die("unable to read files to diff");
1271
1272        if (!DIFF_OPT_TST(o, TEXT) &&
1273            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1274                /* Quite common confusing case */
1275                if (mf1.size == mf2.size &&
1276                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1277                        goto free_ab_and_return;
1278                if (DIFF_OPT_TST(o, BINARY))
1279                        emit_binary_diff(&mf1, &mf2);
1280                else
1281                        printf("Binary files %s and %s differ\n",
1282                               lbl[0], lbl[1]);
1283                o->found_changes = 1;
1284        }
1285        else {
1286                /* Crazy xdl interfaces.. */
1287                const char *diffopts = getenv("GIT_DIFF_OPTS");
1288                xpparam_t xpp;
1289                xdemitconf_t xecfg;
1290                xdemitcb_t ecb;
1291                struct emit_callback ecbdata;
1292                const char *funcname_pattern;
1293
1294                funcname_pattern = diff_funcname_pattern(one);
1295                if (!funcname_pattern)
1296                        funcname_pattern = diff_funcname_pattern(two);
1297
1298                memset(&xecfg, 0, sizeof(xecfg));
1299                memset(&ecbdata, 0, sizeof(ecbdata));
1300                ecbdata.label_path = lbl;
1301                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1302                ecbdata.found_changesp = &o->found_changes;
1303                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1304                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1305                xecfg.ctxlen = o->context;
1306                xecfg.flags = XDL_EMIT_FUNCNAMES;
1307                if (funcname_pattern)
1308                        xdiff_set_find_func(&xecfg, funcname_pattern);
1309                if (!diffopts)
1310                        ;
1311                else if (!prefixcmp(diffopts, "--unified="))
1312                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1313                else if (!prefixcmp(diffopts, "-u"))
1314                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1315                ecb.outf = xdiff_outf;
1316                ecb.priv = &ecbdata;
1317                ecbdata.xm.consume = fn_out_consume;
1318                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1319                        ecbdata.diff_words =
1320                                xcalloc(1, sizeof(struct diff_words_data));
1321                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1322                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1323                        free_diff_words_data(&ecbdata);
1324        }
1325
1326 free_ab_and_return:
1327        diff_free_filespec_data(one);
1328        diff_free_filespec_data(two);
1329        free(a_one);
1330        free(b_two);
1331        return;
1332}
1333
1334static void builtin_diffstat(const char *name_a, const char *name_b,
1335                             struct diff_filespec *one,
1336                             struct diff_filespec *two,
1337                             struct diffstat_t *diffstat,
1338                             struct diff_options *o,
1339                             int complete_rewrite)
1340{
1341        mmfile_t mf1, mf2;
1342        struct diffstat_file *data;
1343
1344        data = diffstat_add(diffstat, name_a, name_b);
1345
1346        if (!one || !two) {
1347                data->is_unmerged = 1;
1348                return;
1349        }
1350        if (complete_rewrite) {
1351                diff_populate_filespec(one, 0);
1352                diff_populate_filespec(two, 0);
1353                data->deleted = count_lines(one->data, one->size);
1354                data->added = count_lines(two->data, two->size);
1355                goto free_and_return;
1356        }
1357        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1358                die("unable to read files to diff");
1359
1360        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1361                data->is_binary = 1;
1362                data->added = mf2.size;
1363                data->deleted = mf1.size;
1364        } else {
1365                /* Crazy xdl interfaces.. */
1366                xpparam_t xpp;
1367                xdemitconf_t xecfg;
1368                xdemitcb_t ecb;
1369
1370                memset(&xecfg, 0, sizeof(xecfg));
1371                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1372                ecb.outf = xdiff_outf;
1373                ecb.priv = diffstat;
1374                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1375        }
1376
1377 free_and_return:
1378        diff_free_filespec_data(one);
1379        diff_free_filespec_data(two);
1380}
1381
1382static void builtin_checkdiff(const char *name_a, const char *name_b,
1383                             struct diff_filespec *one,
1384                             struct diff_filespec *two, struct diff_options *o)
1385{
1386        mmfile_t mf1, mf2;
1387        struct checkdiff_t data;
1388
1389        if (!two)
1390                return;
1391
1392        memset(&data, 0, sizeof(data));
1393        data.xm.consume = checkdiff_consume;
1394        data.filename = name_b ? name_b : name_a;
1395        data.lineno = 0;
1396        data.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1397        data.ws_rule = whitespace_rule(data.filename);
1398
1399        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1400                die("unable to read files to diff");
1401
1402        if (diff_filespec_is_binary(two))
1403                goto free_and_return;
1404        else {
1405                /* Crazy xdl interfaces.. */
1406                xpparam_t xpp;
1407                xdemitconf_t xecfg;
1408                xdemitcb_t ecb;
1409
1410                memset(&xecfg, 0, sizeof(xecfg));
1411                xpp.flags = XDF_NEED_MINIMAL;
1412                ecb.outf = xdiff_outf;
1413                ecb.priv = &data;
1414                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1415        }
1416 free_and_return:
1417        diff_free_filespec_data(one);
1418        diff_free_filespec_data(two);
1419        if (data.status)
1420                DIFF_OPT_SET(o, CHECK_FAILED);
1421}
1422
1423struct diff_filespec *alloc_filespec(const char *path)
1424{
1425        int namelen = strlen(path);
1426        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1427
1428        memset(spec, 0, sizeof(*spec));
1429        spec->path = (char *)(spec + 1);
1430        memcpy(spec->path, path, namelen+1);
1431        spec->count = 1;
1432        return spec;
1433}
1434
1435void free_filespec(struct diff_filespec *spec)
1436{
1437        if (!--spec->count) {
1438                diff_free_filespec_data(spec);
1439                free(spec);
1440        }
1441}
1442
1443void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1444                   unsigned short mode)
1445{
1446        if (mode) {
1447                spec->mode = canon_mode(mode);
1448                hashcpy(spec->sha1, sha1);
1449                spec->sha1_valid = !is_null_sha1(sha1);
1450        }
1451}
1452
1453/*
1454 * Given a name and sha1 pair, if the index tells us the file in
1455 * the work tree has that object contents, return true, so that
1456 * prepare_temp_file() does not have to inflate and extract.
1457 */
1458static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1459{
1460        struct cache_entry *ce;
1461        struct stat st;
1462        int pos, len;
1463
1464        /* We do not read the cache ourselves here, because the
1465         * benchmark with my previous version that always reads cache
1466         * shows that it makes things worse for diff-tree comparing
1467         * two linux-2.6 kernel trees in an already checked out work
1468         * tree.  This is because most diff-tree comparisons deal with
1469         * only a small number of files, while reading the cache is
1470         * expensive for a large project, and its cost outweighs the
1471         * savings we get by not inflating the object to a temporary
1472         * file.  Practically, this code only helps when we are used
1473         * by diff-cache --cached, which does read the cache before
1474         * calling us.
1475         */
1476        if (!active_cache)
1477                return 0;
1478
1479        /* We want to avoid the working directory if our caller
1480         * doesn't need the data in a normal file, this system
1481         * is rather slow with its stat/open/mmap/close syscalls,
1482         * and the object is contained in a pack file.  The pack
1483         * is probably already open and will be faster to obtain
1484         * the data through than the working directory.  Loose
1485         * objects however would tend to be slower as they need
1486         * to be individually opened and inflated.
1487         */
1488        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1489                return 0;
1490
1491        len = strlen(name);
1492        pos = cache_name_pos(name, len);
1493        if (pos < 0)
1494                return 0;
1495        ce = active_cache[pos];
1496        if ((lstat(name, &st) < 0) ||
1497            !S_ISREG(st.st_mode) || /* careful! */
1498            ce_match_stat(ce, &st, 0) ||
1499            hashcmp(sha1, ce->sha1))
1500                return 0;
1501        /* we return 1 only when we can stat, it is a regular file,
1502         * stat information matches, and sha1 recorded in the cache
1503         * matches.  I.e. we know the file in the work tree really is
1504         * the same as the <name, sha1> pair.
1505         */
1506        return 1;
1507}
1508
1509static int populate_from_stdin(struct diff_filespec *s)
1510{
1511        struct strbuf buf;
1512        size_t size = 0;
1513
1514        strbuf_init(&buf, 0);
1515        if (strbuf_read(&buf, 0, 0) < 0)
1516                return error("error while reading from stdin %s",
1517                                     strerror(errno));
1518
1519        s->should_munmap = 0;
1520        s->data = strbuf_detach(&buf, &size);
1521        s->size = size;
1522        s->should_free = 1;
1523        return 0;
1524}
1525
1526static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1527{
1528        int len;
1529        char *data = xmalloc(100);
1530        len = snprintf(data, 100,
1531                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1532        s->data = data;
1533        s->size = len;
1534        s->should_free = 1;
1535        if (size_only) {
1536                s->data = NULL;
1537                free(data);
1538        }
1539        return 0;
1540}
1541
1542/*
1543 * While doing rename detection and pickaxe operation, we may need to
1544 * grab the data for the blob (or file) for our own in-core comparison.
1545 * diff_filespec has data and size fields for this purpose.
1546 */
1547int diff_populate_filespec(struct diff_filespec *s, int size_only)
1548{
1549        int err = 0;
1550        if (!DIFF_FILE_VALID(s))
1551                die("internal error: asking to populate invalid file.");
1552        if (S_ISDIR(s->mode))
1553                return -1;
1554
1555        if (s->data)
1556                return 0;
1557
1558        if (size_only && 0 < s->size)
1559                return 0;
1560
1561        if (S_ISGITLINK(s->mode))
1562                return diff_populate_gitlink(s, size_only);
1563
1564        if (!s->sha1_valid ||
1565            reuse_worktree_file(s->path, s->sha1, 0)) {
1566                struct strbuf buf;
1567                struct stat st;
1568                int fd;
1569
1570                if (!strcmp(s->path, "-"))
1571                        return populate_from_stdin(s);
1572
1573                if (lstat(s->path, &st) < 0) {
1574                        if (errno == ENOENT) {
1575                        err_empty:
1576                                err = -1;
1577                        empty:
1578                                s->data = (char *)"";
1579                                s->size = 0;
1580                                return err;
1581                        }
1582                }
1583                s->size = xsize_t(st.st_size);
1584                if (!s->size)
1585                        goto empty;
1586                if (size_only)
1587                        return 0;
1588                if (S_ISLNK(st.st_mode)) {
1589                        int ret;
1590                        s->data = xmalloc(s->size);
1591                        s->should_free = 1;
1592                        ret = readlink(s->path, s->data, s->size);
1593                        if (ret < 0) {
1594                                free(s->data);
1595                                goto err_empty;
1596                        }
1597                        return 0;
1598                }
1599                fd = open(s->path, O_RDONLY);
1600                if (fd < 0)
1601                        goto err_empty;
1602                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1603                close(fd);
1604                s->should_munmap = 1;
1605
1606                /*
1607                 * Convert from working tree format to canonical git format
1608                 */
1609                strbuf_init(&buf, 0);
1610                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1611                        size_t size = 0;
1612                        munmap(s->data, s->size);
1613                        s->should_munmap = 0;
1614                        s->data = strbuf_detach(&buf, &size);
1615                        s->size = size;
1616                        s->should_free = 1;
1617                }
1618        }
1619        else {
1620                enum object_type type;
1621                if (size_only)
1622                        type = sha1_object_info(s->sha1, &s->size);
1623                else {
1624                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1625                        s->should_free = 1;
1626                }
1627        }
1628        return 0;
1629}
1630
1631void diff_free_filespec_blob(struct diff_filespec *s)
1632{
1633        if (s->should_free)
1634                free(s->data);
1635        else if (s->should_munmap)
1636                munmap(s->data, s->size);
1637
1638        if (s->should_free || s->should_munmap) {
1639                s->should_free = s->should_munmap = 0;
1640                s->data = NULL;
1641        }
1642}
1643
1644void diff_free_filespec_data(struct diff_filespec *s)
1645{
1646        diff_free_filespec_blob(s);
1647        free(s->cnt_data);
1648        s->cnt_data = NULL;
1649}
1650
1651static void prep_temp_blob(struct diff_tempfile *temp,
1652                           void *blob,
1653                           unsigned long size,
1654                           const unsigned char *sha1,
1655                           int mode)
1656{
1657        int fd;
1658
1659        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1660        if (fd < 0)
1661                die("unable to create temp-file: %s", strerror(errno));
1662        if (write_in_full(fd, blob, size) != size)
1663                die("unable to write temp-file");
1664        close(fd);
1665        temp->name = temp->tmp_path;
1666        strcpy(temp->hex, sha1_to_hex(sha1));
1667        temp->hex[40] = 0;
1668        sprintf(temp->mode, "%06o", mode);
1669}
1670
1671static void prepare_temp_file(const char *name,
1672                              struct diff_tempfile *temp,
1673                              struct diff_filespec *one)
1674{
1675        if (!DIFF_FILE_VALID(one)) {
1676        not_a_valid_file:
1677                /* A '-' entry produces this for file-2, and
1678                 * a '+' entry produces this for file-1.
1679                 */
1680                temp->name = "/dev/null";
1681                strcpy(temp->hex, ".");
1682                strcpy(temp->mode, ".");
1683                return;
1684        }
1685
1686        if (!one->sha1_valid ||
1687            reuse_worktree_file(name, one->sha1, 1)) {
1688                struct stat st;
1689                if (lstat(name, &st) < 0) {
1690                        if (errno == ENOENT)
1691                                goto not_a_valid_file;
1692                        die("stat(%s): %s", name, strerror(errno));
1693                }
1694                if (S_ISLNK(st.st_mode)) {
1695                        int ret;
1696                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1697                        size_t sz = xsize_t(st.st_size);
1698                        if (sizeof(buf) <= st.st_size)
1699                                die("symlink too long: %s", name);
1700                        ret = readlink(name, buf, sz);
1701                        if (ret < 0)
1702                                die("readlink(%s)", name);
1703                        prep_temp_blob(temp, buf, sz,
1704                                       (one->sha1_valid ?
1705                                        one->sha1 : null_sha1),
1706                                       (one->sha1_valid ?
1707                                        one->mode : S_IFLNK));
1708                }
1709                else {
1710                        /* we can borrow from the file in the work tree */
1711                        temp->name = name;
1712                        if (!one->sha1_valid)
1713                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1714                        else
1715                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1716                        /* Even though we may sometimes borrow the
1717                         * contents from the work tree, we always want
1718                         * one->mode.  mode is trustworthy even when
1719                         * !(one->sha1_valid), as long as
1720                         * DIFF_FILE_VALID(one).
1721                         */
1722                        sprintf(temp->mode, "%06o", one->mode);
1723                }
1724                return;
1725        }
1726        else {
1727                if (diff_populate_filespec(one, 0))
1728                        die("cannot read data blob for %s", one->path);
1729                prep_temp_blob(temp, one->data, one->size,
1730                               one->sha1, one->mode);
1731        }
1732}
1733
1734static void remove_tempfile(void)
1735{
1736        int i;
1737
1738        for (i = 0; i < 2; i++)
1739                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1740                        unlink(diff_temp[i].name);
1741                        diff_temp[i].name = NULL;
1742                }
1743}
1744
1745static void remove_tempfile_on_signal(int signo)
1746{
1747        remove_tempfile();
1748        signal(SIGINT, SIG_DFL);
1749        raise(signo);
1750}
1751
1752/* An external diff command takes:
1753 *
1754 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1755 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1756 *
1757 */
1758static void run_external_diff(const char *pgm,
1759                              const char *name,
1760                              const char *other,
1761                              struct diff_filespec *one,
1762                              struct diff_filespec *two,
1763                              const char *xfrm_msg,
1764                              int complete_rewrite)
1765{
1766        const char *spawn_arg[10];
1767        struct diff_tempfile *temp = diff_temp;
1768        int retval;
1769        static int atexit_asked = 0;
1770        const char *othername;
1771        const char **arg = &spawn_arg[0];
1772
1773        othername = (other? other : name);
1774        if (one && two) {
1775                prepare_temp_file(name, &temp[0], one);
1776                prepare_temp_file(othername, &temp[1], two);
1777                if (! atexit_asked &&
1778                    (temp[0].name == temp[0].tmp_path ||
1779                     temp[1].name == temp[1].tmp_path)) {
1780                        atexit_asked = 1;
1781                        atexit(remove_tempfile);
1782                }
1783                signal(SIGINT, remove_tempfile_on_signal);
1784        }
1785
1786        if (one && two) {
1787                *arg++ = pgm;
1788                *arg++ = name;
1789                *arg++ = temp[0].name;
1790                *arg++ = temp[0].hex;
1791                *arg++ = temp[0].mode;
1792                *arg++ = temp[1].name;
1793                *arg++ = temp[1].hex;
1794                *arg++ = temp[1].mode;
1795                if (other) {
1796                        *arg++ = other;
1797                        *arg++ = xfrm_msg;
1798                }
1799        } else {
1800                *arg++ = pgm;
1801                *arg++ = name;
1802        }
1803        *arg = NULL;
1804        fflush(NULL);
1805        retval = run_command_v_opt(spawn_arg, 0);
1806        remove_tempfile();
1807        if (retval) {
1808                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1809                exit(1);
1810        }
1811}
1812
1813static const char *external_diff_attr(const char *name)
1814{
1815        struct git_attr_check attr_diff_check;
1816
1817        setup_diff_attr_check(&attr_diff_check);
1818        if (!git_checkattr(name, 1, &attr_diff_check)) {
1819                const char *value = attr_diff_check.value;
1820                if (!ATTR_TRUE(value) &&
1821                    !ATTR_FALSE(value) &&
1822                    !ATTR_UNSET(value)) {
1823                        struct ll_diff_driver *drv;
1824
1825                        read_config_if_needed();
1826                        for (drv = user_diff; drv; drv = drv->next)
1827                                if (!strcmp(drv->name, value))
1828                                        return drv->cmd;
1829                }
1830        }
1831        return NULL;
1832}
1833
1834static void run_diff_cmd(const char *pgm,
1835                         const char *name,
1836                         const char *other,
1837                         struct diff_filespec *one,
1838                         struct diff_filespec *two,
1839                         const char *xfrm_msg,
1840                         struct diff_options *o,
1841                         int complete_rewrite)
1842{
1843        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
1844                pgm = NULL;
1845        else {
1846                const char *cmd = external_diff_attr(name);
1847                if (cmd)
1848                        pgm = cmd;
1849        }
1850
1851        if (pgm) {
1852                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1853                                  complete_rewrite);
1854                return;
1855        }
1856        if (one && two)
1857                builtin_diff(name, other ? other : name,
1858                             one, two, xfrm_msg, o, complete_rewrite);
1859        else
1860                printf("* Unmerged path %s\n", name);
1861}
1862
1863static void diff_fill_sha1_info(struct diff_filespec *one)
1864{
1865        if (DIFF_FILE_VALID(one)) {
1866                if (!one->sha1_valid) {
1867                        struct stat st;
1868                        if (!strcmp(one->path, "-")) {
1869                                hashcpy(one->sha1, null_sha1);
1870                                return;
1871                        }
1872                        if (lstat(one->path, &st) < 0)
1873                                die("stat %s", one->path);
1874                        if (index_path(one->sha1, one->path, &st, 0))
1875                                die("cannot hash %s\n", one->path);
1876                }
1877        }
1878        else
1879                hashclr(one->sha1);
1880}
1881
1882static int similarity_index(struct diff_filepair *p)
1883{
1884        return p->score * 100 / MAX_SCORE;
1885}
1886
1887static void run_diff(struct diff_filepair *p, struct diff_options *o)
1888{
1889        const char *pgm = external_diff();
1890        struct strbuf msg;
1891        char *xfrm_msg;
1892        struct diff_filespec *one = p->one;
1893        struct diff_filespec *two = p->two;
1894        const char *name;
1895        const char *other;
1896        int complete_rewrite = 0;
1897
1898
1899        if (DIFF_PAIR_UNMERGED(p)) {
1900                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1901                return;
1902        }
1903
1904        name  = p->one->path;
1905        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1906        diff_fill_sha1_info(one);
1907        diff_fill_sha1_info(two);
1908
1909        strbuf_init(&msg, PATH_MAX * 2 + 300);
1910        switch (p->status) {
1911        case DIFF_STATUS_COPIED:
1912                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1913                strbuf_addstr(&msg, "\ncopy from ");
1914                quote_c_style(name, &msg, NULL, 0);
1915                strbuf_addstr(&msg, "\ncopy to ");
1916                quote_c_style(other, &msg, NULL, 0);
1917                strbuf_addch(&msg, '\n');
1918                break;
1919        case DIFF_STATUS_RENAMED:
1920                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1921                strbuf_addstr(&msg, "\nrename from ");
1922                quote_c_style(name, &msg, NULL, 0);
1923                strbuf_addstr(&msg, "\nrename to ");
1924                quote_c_style(other, &msg, NULL, 0);
1925                strbuf_addch(&msg, '\n');
1926                break;
1927        case DIFF_STATUS_MODIFIED:
1928                if (p->score) {
1929                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
1930                                        similarity_index(p));
1931                        complete_rewrite = 1;
1932                        break;
1933                }
1934                /* fallthru */
1935        default:
1936                /* nothing */
1937                ;
1938        }
1939
1940        if (hashcmp(one->sha1, two->sha1)) {
1941                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
1942
1943                if (DIFF_OPT_TST(o, BINARY)) {
1944                        mmfile_t mf;
1945                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
1946                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
1947                                abbrev = 40;
1948                }
1949                strbuf_addf(&msg, "index %.*s..%.*s",
1950                                abbrev, sha1_to_hex(one->sha1),
1951                                abbrev, sha1_to_hex(two->sha1));
1952                if (one->mode == two->mode)
1953                        strbuf_addf(&msg, " %06o", one->mode);
1954                strbuf_addch(&msg, '\n');
1955        }
1956
1957        if (msg.len)
1958                strbuf_setlen(&msg, msg.len - 1);
1959        xfrm_msg = msg.len ? msg.buf : NULL;
1960
1961        if (!pgm &&
1962            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1963            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1964                /* a filepair that changes between file and symlink
1965                 * needs to be split into deletion and creation.
1966                 */
1967                struct diff_filespec *null = alloc_filespec(two->path);
1968                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1969                free(null);
1970                null = alloc_filespec(one->path);
1971                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1972                free(null);
1973        }
1974        else
1975                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
1976                             complete_rewrite);
1977
1978        strbuf_release(&msg);
1979}
1980
1981static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
1982                         struct diffstat_t *diffstat)
1983{
1984        const char *name;
1985        const char *other;
1986        int complete_rewrite = 0;
1987
1988        if (DIFF_PAIR_UNMERGED(p)) {
1989                /* unmerged */
1990                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
1991                return;
1992        }
1993
1994        name = p->one->path;
1995        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1996
1997        diff_fill_sha1_info(p->one);
1998        diff_fill_sha1_info(p->two);
1999
2000        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2001                complete_rewrite = 1;
2002        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2003}
2004
2005static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2006{
2007        const char *name;
2008        const char *other;
2009
2010        if (DIFF_PAIR_UNMERGED(p)) {
2011                /* unmerged */
2012                return;
2013        }
2014
2015        name = p->one->path;
2016        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2017
2018        diff_fill_sha1_info(p->one);
2019        diff_fill_sha1_info(p->two);
2020
2021        builtin_checkdiff(name, other, p->one, p->two, o);
2022}
2023
2024void diff_setup(struct diff_options *options)
2025{
2026        memset(options, 0, sizeof(*options));
2027        options->line_termination = '\n';
2028        options->break_opt = -1;
2029        options->rename_limit = -1;
2030        options->context = 3;
2031        options->msg_sep = "";
2032
2033        options->change = diff_change;
2034        options->add_remove = diff_addremove;
2035        if (diff_use_color_default)
2036                DIFF_OPT_SET(options, COLOR_DIFF);
2037        else
2038                DIFF_OPT_CLR(options, COLOR_DIFF);
2039        options->detect_rename = diff_detect_rename_default;
2040
2041        options->a_prefix = "a/";
2042        options->b_prefix = "b/";
2043}
2044
2045int diff_setup_done(struct diff_options *options)
2046{
2047        int count = 0;
2048
2049        if (options->output_format & DIFF_FORMAT_NAME)
2050                count++;
2051        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2052                count++;
2053        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2054                count++;
2055        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2056                count++;
2057        if (count > 1)
2058                die("--name-only, --name-status, --check and -s are mutually exclusive");
2059
2060        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2061                options->detect_rename = DIFF_DETECT_COPY;
2062
2063        if (options->output_format & (DIFF_FORMAT_NAME |
2064                                      DIFF_FORMAT_NAME_STATUS |
2065                                      DIFF_FORMAT_CHECKDIFF |
2066                                      DIFF_FORMAT_NO_OUTPUT))
2067                options->output_format &= ~(DIFF_FORMAT_RAW |
2068                                            DIFF_FORMAT_NUMSTAT |
2069                                            DIFF_FORMAT_DIFFSTAT |
2070                                            DIFF_FORMAT_SHORTSTAT |
2071                                            DIFF_FORMAT_SUMMARY |
2072                                            DIFF_FORMAT_PATCH);
2073
2074        /*
2075         * These cases always need recursive; we do not drop caller-supplied
2076         * recursive bits for other formats here.
2077         */
2078        if (options->output_format & (DIFF_FORMAT_PATCH |
2079                                      DIFF_FORMAT_NUMSTAT |
2080                                      DIFF_FORMAT_DIFFSTAT |
2081                                      DIFF_FORMAT_SHORTSTAT |
2082                                      DIFF_FORMAT_SUMMARY |
2083                                      DIFF_FORMAT_CHECKDIFF))
2084                DIFF_OPT_SET(options, RECURSIVE);
2085        /*
2086         * Also pickaxe would not work very well if you do not say recursive
2087         */
2088        if (options->pickaxe)
2089                DIFF_OPT_SET(options, RECURSIVE);
2090
2091        if (options->detect_rename && options->rename_limit < 0)
2092                options->rename_limit = diff_rename_limit_default;
2093        if (options->setup & DIFF_SETUP_USE_CACHE) {
2094                if (!active_cache)
2095                        /* read-cache does not die even when it fails
2096                         * so it is safe for us to do this here.  Also
2097                         * it does not smudge active_cache or active_nr
2098                         * when it fails, so we do not have to worry about
2099                         * cleaning it up ourselves either.
2100                         */
2101                        read_cache();
2102        }
2103        if (options->abbrev <= 0 || 40 < options->abbrev)
2104                options->abbrev = 40; /* full */
2105
2106        /*
2107         * It does not make sense to show the first hit we happened
2108         * to have found.  It does not make sense not to return with
2109         * exit code in such a case either.
2110         */
2111        if (DIFF_OPT_TST(options, QUIET)) {
2112                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2113                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2114        }
2115
2116        /*
2117         * If we postprocess in diffcore, we cannot simply return
2118         * upon the first hit.  We need to run diff as usual.
2119         */
2120        if (options->pickaxe || options->filter)
2121                DIFF_OPT_CLR(options, QUIET);
2122
2123        return 0;
2124}
2125
2126static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2127{
2128        char c, *eq;
2129        int len;
2130
2131        if (*arg != '-')
2132                return 0;
2133        c = *++arg;
2134        if (!c)
2135                return 0;
2136        if (c == arg_short) {
2137                c = *++arg;
2138                if (!c)
2139                        return 1;
2140                if (val && isdigit(c)) {
2141                        char *end;
2142                        int n = strtoul(arg, &end, 10);
2143                        if (*end)
2144                                return 0;
2145                        *val = n;
2146                        return 1;
2147                }
2148                return 0;
2149        }
2150        if (c != '-')
2151                return 0;
2152        arg++;
2153        eq = strchr(arg, '=');
2154        if (eq)
2155                len = eq - arg;
2156        else
2157                len = strlen(arg);
2158        if (!len || strncmp(arg, arg_long, len))
2159                return 0;
2160        if (eq) {
2161                int n;
2162                char *end;
2163                if (!isdigit(*++eq))
2164                        return 0;
2165                n = strtoul(eq, &end, 10);
2166                if (*end)
2167                        return 0;
2168                *val = n;
2169        }
2170        return 1;
2171}
2172
2173static int diff_scoreopt_parse(const char *opt);
2174
2175int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2176{
2177        const char *arg = av[0];
2178
2179        /* Output format options */
2180        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2181                options->output_format |= DIFF_FORMAT_PATCH;
2182        else if (opt_arg(arg, 'U', "unified", &options->context))
2183                options->output_format |= DIFF_FORMAT_PATCH;
2184        else if (!strcmp(arg, "--raw"))
2185                options->output_format |= DIFF_FORMAT_RAW;
2186        else if (!strcmp(arg, "--patch-with-raw"))
2187                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2188        else if (!strcmp(arg, "--numstat"))
2189                options->output_format |= DIFF_FORMAT_NUMSTAT;
2190        else if (!strcmp(arg, "--shortstat"))
2191                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2192        else if (!strcmp(arg, "--check"))
2193                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2194        else if (!strcmp(arg, "--summary"))
2195                options->output_format |= DIFF_FORMAT_SUMMARY;
2196        else if (!strcmp(arg, "--patch-with-stat"))
2197                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2198        else if (!strcmp(arg, "--name-only"))
2199                options->output_format |= DIFF_FORMAT_NAME;
2200        else if (!strcmp(arg, "--name-status"))
2201                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2202        else if (!strcmp(arg, "-s"))
2203                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2204        else if (!prefixcmp(arg, "--stat")) {
2205                char *end;
2206                int width = options->stat_width;
2207                int name_width = options->stat_name_width;
2208                arg += 6;
2209                end = (char *)arg;
2210
2211                switch (*arg) {
2212                case '-':
2213                        if (!prefixcmp(arg, "-width="))
2214                                width = strtoul(arg + 7, &end, 10);
2215                        else if (!prefixcmp(arg, "-name-width="))
2216                                name_width = strtoul(arg + 12, &end, 10);
2217                        break;
2218                case '=':
2219                        width = strtoul(arg+1, &end, 10);
2220                        if (*end == ',')
2221                                name_width = strtoul(end+1, &end, 10);
2222                }
2223
2224                /* Important! This checks all the error cases! */
2225                if (*end)
2226                        return 0;
2227                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2228                options->stat_name_width = name_width;
2229                options->stat_width = width;
2230        }
2231
2232        /* renames options */
2233        else if (!prefixcmp(arg, "-B")) {
2234                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2235                        return -1;
2236        }
2237        else if (!prefixcmp(arg, "-M")) {
2238                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2239                        return -1;
2240                options->detect_rename = DIFF_DETECT_RENAME;
2241        }
2242        else if (!prefixcmp(arg, "-C")) {
2243                if (options->detect_rename == DIFF_DETECT_COPY)
2244                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2245                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2246                        return -1;
2247                options->detect_rename = DIFF_DETECT_COPY;
2248        }
2249        else if (!strcmp(arg, "--no-renames"))
2250                options->detect_rename = 0;
2251
2252        /* xdiff options */
2253        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2254                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2255        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2256                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2257        else if (!strcmp(arg, "--ignore-space-at-eol"))
2258                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2259
2260        /* flags options */
2261        else if (!strcmp(arg, "--binary")) {
2262                options->output_format |= DIFF_FORMAT_PATCH;
2263                DIFF_OPT_SET(options, BINARY);
2264        }
2265        else if (!strcmp(arg, "--full-index"))
2266                DIFF_OPT_SET(options, FULL_INDEX);
2267        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2268                DIFF_OPT_SET(options, TEXT);
2269        else if (!strcmp(arg, "-R"))
2270                DIFF_OPT_SET(options, REVERSE_DIFF);
2271        else if (!strcmp(arg, "--find-copies-harder"))
2272                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2273        else if (!strcmp(arg, "--follow"))
2274                DIFF_OPT_SET(options, FOLLOW_RENAMES);
2275        else if (!strcmp(arg, "--color"))
2276                DIFF_OPT_SET(options, COLOR_DIFF);
2277        else if (!strcmp(arg, "--no-color"))
2278                DIFF_OPT_CLR(options, COLOR_DIFF);
2279        else if (!strcmp(arg, "--color-words"))
2280                options->flags |= DIFF_OPT_COLOR_DIFF | DIFF_OPT_COLOR_DIFF_WORDS;
2281        else if (!strcmp(arg, "--exit-code"))
2282                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2283        else if (!strcmp(arg, "--quiet"))
2284                DIFF_OPT_SET(options, QUIET);
2285        else if (!strcmp(arg, "--ext-diff"))
2286                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2287        else if (!strcmp(arg, "--no-ext-diff"))
2288                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2289
2290        /* misc options */
2291        else if (!strcmp(arg, "-z"))
2292                options->line_termination = 0;
2293        else if (!prefixcmp(arg, "-l"))
2294                options->rename_limit = strtoul(arg+2, NULL, 10);
2295        else if (!prefixcmp(arg, "-S"))
2296                options->pickaxe = arg + 2;
2297        else if (!strcmp(arg, "--pickaxe-all"))
2298                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2299        else if (!strcmp(arg, "--pickaxe-regex"))
2300                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2301        else if (!prefixcmp(arg, "-O"))
2302                options->orderfile = arg + 2;
2303        else if (!prefixcmp(arg, "--diff-filter="))
2304                options->filter = arg + 14;
2305        else if (!strcmp(arg, "--abbrev"))
2306                options->abbrev = DEFAULT_ABBREV;
2307        else if (!prefixcmp(arg, "--abbrev=")) {
2308                options->abbrev = strtoul(arg + 9, NULL, 10);
2309                if (options->abbrev < MINIMUM_ABBREV)
2310                        options->abbrev = MINIMUM_ABBREV;
2311                else if (40 < options->abbrev)
2312                        options->abbrev = 40;
2313        }
2314        else if (!prefixcmp(arg, "--src-prefix="))
2315                options->a_prefix = arg + 13;
2316        else if (!prefixcmp(arg, "--dst-prefix="))
2317                options->b_prefix = arg + 13;
2318        else if (!strcmp(arg, "--no-prefix"))
2319                options->a_prefix = options->b_prefix = "";
2320        else
2321                return 0;
2322        return 1;
2323}
2324
2325static int parse_num(const char **cp_p)
2326{
2327        unsigned long num, scale;
2328        int ch, dot;
2329        const char *cp = *cp_p;
2330
2331        num = 0;
2332        scale = 1;
2333        dot = 0;
2334        for(;;) {
2335                ch = *cp;
2336                if ( !dot && ch == '.' ) {
2337                        scale = 1;
2338                        dot = 1;
2339                } else if ( ch == '%' ) {
2340                        scale = dot ? scale*100 : 100;
2341                        cp++;   /* % is always at the end */
2342                        break;
2343                } else if ( ch >= '0' && ch <= '9' ) {
2344                        if ( scale < 100000 ) {
2345                                scale *= 10;
2346                                num = (num*10) + (ch-'0');
2347                        }
2348                } else {
2349                        break;
2350                }
2351                cp++;
2352        }
2353        *cp_p = cp;
2354
2355        /* user says num divided by scale and we say internally that
2356         * is MAX_SCORE * num / scale.
2357         */
2358        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2359}
2360
2361static int diff_scoreopt_parse(const char *opt)
2362{
2363        int opt1, opt2, cmd;
2364
2365        if (*opt++ != '-')
2366                return -1;
2367        cmd = *opt++;
2368        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2369                return -1; /* that is not a -M, -C nor -B option */
2370
2371        opt1 = parse_num(&opt);
2372        if (cmd != 'B')
2373                opt2 = 0;
2374        else {
2375                if (*opt == 0)
2376                        opt2 = 0;
2377                else if (*opt != '/')
2378                        return -1; /* we expect -B80/99 or -B80 */
2379                else {
2380                        opt++;
2381                        opt2 = parse_num(&opt);
2382                }
2383        }
2384        if (*opt != 0)
2385                return -1;
2386        return opt1 | (opt2 << 16);
2387}
2388
2389struct diff_queue_struct diff_queued_diff;
2390
2391void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2392{
2393        if (queue->alloc <= queue->nr) {
2394                queue->alloc = alloc_nr(queue->alloc);
2395                queue->queue = xrealloc(queue->queue,
2396                                        sizeof(dp) * queue->alloc);
2397        }
2398        queue->queue[queue->nr++] = dp;
2399}
2400
2401struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2402                                 struct diff_filespec *one,
2403                                 struct diff_filespec *two)
2404{
2405        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2406        dp->one = one;
2407        dp->two = two;
2408        if (queue)
2409                diff_q(queue, dp);
2410        return dp;
2411}
2412
2413void diff_free_filepair(struct diff_filepair *p)
2414{
2415        free_filespec(p->one);
2416        free_filespec(p->two);
2417        free(p);
2418}
2419
2420/* This is different from find_unique_abbrev() in that
2421 * it stuffs the result with dots for alignment.
2422 */
2423const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2424{
2425        int abblen;
2426        const char *abbrev;
2427        if (len == 40)
2428                return sha1_to_hex(sha1);
2429
2430        abbrev = find_unique_abbrev(sha1, len);
2431        if (!abbrev)
2432                return sha1_to_hex(sha1);
2433        abblen = strlen(abbrev);
2434        if (abblen < 37) {
2435                static char hex[41];
2436                if (len < abblen && abblen <= len + 2)
2437                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2438                else
2439                        sprintf(hex, "%s...", abbrev);
2440                return hex;
2441        }
2442        return sha1_to_hex(sha1);
2443}
2444
2445static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2446{
2447        int line_termination = opt->line_termination;
2448        int inter_name_termination = line_termination ? '\t' : '\0';
2449
2450        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2451                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2452                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2453                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2454        }
2455        if (p->score) {
2456                printf("%c%03d%c", p->status, similarity_index(p),
2457                           inter_name_termination);
2458        } else {
2459                printf("%c%c", p->status, inter_name_termination);
2460        }
2461
2462        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2463                write_name_quoted(p->one->path, stdout, inter_name_termination);
2464                write_name_quoted(p->two->path, stdout, line_termination);
2465        } else {
2466                const char *path = p->one->mode ? p->one->path : p->two->path;
2467                write_name_quoted(path, stdout, line_termination);
2468        }
2469}
2470
2471int diff_unmodified_pair(struct diff_filepair *p)
2472{
2473        /* This function is written stricter than necessary to support
2474         * the currently implemented transformers, but the idea is to
2475         * let transformers to produce diff_filepairs any way they want,
2476         * and filter and clean them up here before producing the output.
2477         */
2478        struct diff_filespec *one = p->one, *two = p->two;
2479
2480        if (DIFF_PAIR_UNMERGED(p))
2481                return 0; /* unmerged is interesting */
2482
2483        /* deletion, addition, mode or type change
2484         * and rename are all interesting.
2485         */
2486        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2487            DIFF_PAIR_MODE_CHANGED(p) ||
2488            strcmp(one->path, two->path))
2489                return 0;
2490
2491        /* both are valid and point at the same path.  that is, we are
2492         * dealing with a change.
2493         */
2494        if (one->sha1_valid && two->sha1_valid &&
2495            !hashcmp(one->sha1, two->sha1))
2496                return 1; /* no change */
2497        if (!one->sha1_valid && !two->sha1_valid)
2498                return 1; /* both look at the same file on the filesystem. */
2499        return 0;
2500}
2501
2502static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2503{
2504        if (diff_unmodified_pair(p))
2505                return;
2506
2507        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2508            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2509                return; /* no tree diffs in patch format */
2510
2511        run_diff(p, o);
2512}
2513
2514static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2515                            struct diffstat_t *diffstat)
2516{
2517        if (diff_unmodified_pair(p))
2518                return;
2519
2520        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2521            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2522                return; /* no tree diffs in patch format */
2523
2524        run_diffstat(p, o, diffstat);
2525}
2526
2527static void diff_flush_checkdiff(struct diff_filepair *p,
2528                struct diff_options *o)
2529{
2530        if (diff_unmodified_pair(p))
2531                return;
2532
2533        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2534            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2535                return; /* no tree diffs in patch format */
2536
2537        run_checkdiff(p, o);
2538}
2539
2540int diff_queue_is_empty(void)
2541{
2542        struct diff_queue_struct *q = &diff_queued_diff;
2543        int i;
2544        for (i = 0; i < q->nr; i++)
2545                if (!diff_unmodified_pair(q->queue[i]))
2546                        return 0;
2547        return 1;
2548}
2549
2550#if DIFF_DEBUG
2551void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2552{
2553        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2554                x, one ? one : "",
2555                s->path,
2556                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2557                s->mode,
2558                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2559        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2560                x, one ? one : "",
2561                s->size, s->xfrm_flags);
2562}
2563
2564void diff_debug_filepair(const struct diff_filepair *p, int i)
2565{
2566        diff_debug_filespec(p->one, i, "one");
2567        diff_debug_filespec(p->two, i, "two");
2568        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2569                p->score, p->status ? p->status : '?',
2570                p->one->rename_used, p->broken_pair);
2571}
2572
2573void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2574{
2575        int i;
2576        if (msg)
2577                fprintf(stderr, "%s\n", msg);
2578        fprintf(stderr, "q->nr = %d\n", q->nr);
2579        for (i = 0; i < q->nr; i++) {
2580                struct diff_filepair *p = q->queue[i];
2581                diff_debug_filepair(p, i);
2582        }
2583}
2584#endif
2585
2586static void diff_resolve_rename_copy(void)
2587{
2588        int i;
2589        struct diff_filepair *p;
2590        struct diff_queue_struct *q = &diff_queued_diff;
2591
2592        diff_debug_queue("resolve-rename-copy", q);
2593
2594        for (i = 0; i < q->nr; i++) {
2595                p = q->queue[i];
2596                p->status = 0; /* undecided */
2597                if (DIFF_PAIR_UNMERGED(p))
2598                        p->status = DIFF_STATUS_UNMERGED;
2599                else if (!DIFF_FILE_VALID(p->one))
2600                        p->status = DIFF_STATUS_ADDED;
2601                else if (!DIFF_FILE_VALID(p->two))
2602                        p->status = DIFF_STATUS_DELETED;
2603                else if (DIFF_PAIR_TYPE_CHANGED(p))
2604                        p->status = DIFF_STATUS_TYPE_CHANGED;
2605
2606                /* from this point on, we are dealing with a pair
2607                 * whose both sides are valid and of the same type, i.e.
2608                 * either in-place edit or rename/copy edit.
2609                 */
2610                else if (DIFF_PAIR_RENAME(p)) {
2611                        /*
2612                         * A rename might have re-connected a broken
2613                         * pair up, causing the pathnames to be the
2614                         * same again. If so, that's not a rename at
2615                         * all, just a modification..
2616                         *
2617                         * Otherwise, see if this source was used for
2618                         * multiple renames, in which case we decrement
2619                         * the count, and call it a copy.
2620                         */
2621                        if (!strcmp(p->one->path, p->two->path))
2622                                p->status = DIFF_STATUS_MODIFIED;
2623                        else if (--p->one->rename_used > 0)
2624                                p->status = DIFF_STATUS_COPIED;
2625                        else
2626                                p->status = DIFF_STATUS_RENAMED;
2627                }
2628                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2629                         p->one->mode != p->two->mode ||
2630                         is_null_sha1(p->one->sha1))
2631                        p->status = DIFF_STATUS_MODIFIED;
2632                else {
2633                        /* This is a "no-change" entry and should not
2634                         * happen anymore, but prepare for broken callers.
2635                         */
2636                        error("feeding unmodified %s to diffcore",
2637                              p->one->path);
2638                        p->status = DIFF_STATUS_UNKNOWN;
2639                }
2640        }
2641        diff_debug_queue("resolve-rename-copy done", q);
2642}
2643
2644static int check_pair_status(struct diff_filepair *p)
2645{
2646        switch (p->status) {
2647        case DIFF_STATUS_UNKNOWN:
2648                return 0;
2649        case 0:
2650                die("internal error in diff-resolve-rename-copy");
2651        default:
2652                return 1;
2653        }
2654}
2655
2656static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2657{
2658        int fmt = opt->output_format;
2659
2660        if (fmt & DIFF_FORMAT_CHECKDIFF)
2661                diff_flush_checkdiff(p, opt);
2662        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2663                diff_flush_raw(p, opt);
2664        else if (fmt & DIFF_FORMAT_NAME)
2665                write_name_quoted(p->two->path, stdout, opt->line_termination);
2666}
2667
2668static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2669{
2670        if (fs->mode)
2671                printf(" %s mode %06o ", newdelete, fs->mode);
2672        else
2673                printf(" %s ", newdelete);
2674        write_name_quoted(fs->path, stdout, '\n');
2675}
2676
2677
2678static void show_mode_change(struct diff_filepair *p, int show_name)
2679{
2680        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2681                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2682                        show_name ? ' ' : '\n');
2683                if (show_name) {
2684                        write_name_quoted(p->two->path, stdout, '\n');
2685                }
2686        }
2687}
2688
2689static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2690{
2691        char *names = pprint_rename(p->one->path, p->two->path);
2692
2693        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2694        free(names);
2695        show_mode_change(p, 0);
2696}
2697
2698static void diff_summary(struct diff_filepair *p)
2699{
2700        switch(p->status) {
2701        case DIFF_STATUS_DELETED:
2702                show_file_mode_name("delete", p->one);
2703                break;
2704        case DIFF_STATUS_ADDED:
2705                show_file_mode_name("create", p->two);
2706                break;
2707        case DIFF_STATUS_COPIED:
2708                show_rename_copy("copy", p);
2709                break;
2710        case DIFF_STATUS_RENAMED:
2711                show_rename_copy("rename", p);
2712                break;
2713        default:
2714                if (p->score) {
2715                        fputs(" rewrite ", stdout);
2716                        write_name_quoted(p->two->path, stdout, ' ');
2717                        printf("(%d%%)\n", similarity_index(p));
2718                }
2719                show_mode_change(p, !p->score);
2720                break;
2721        }
2722}
2723
2724struct patch_id_t {
2725        struct xdiff_emit_state xm;
2726        SHA_CTX *ctx;
2727        int patchlen;
2728};
2729
2730static int remove_space(char *line, int len)
2731{
2732        int i;
2733        char *dst = line;
2734        unsigned char c;
2735
2736        for (i = 0; i < len; i++)
2737                if (!isspace((c = line[i])))
2738                        *dst++ = c;
2739
2740        return dst - line;
2741}
2742
2743static void patch_id_consume(void *priv, char *line, unsigned long len)
2744{
2745        struct patch_id_t *data = priv;
2746        int new_len;
2747
2748        /* Ignore line numbers when computing the SHA1 of the patch */
2749        if (!prefixcmp(line, "@@ -"))
2750                return;
2751
2752        new_len = remove_space(line, len);
2753
2754        SHA1_Update(data->ctx, line, new_len);
2755        data->patchlen += new_len;
2756}
2757
2758/* returns 0 upon success, and writes result into sha1 */
2759static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2760{
2761        struct diff_queue_struct *q = &diff_queued_diff;
2762        int i;
2763        SHA_CTX ctx;
2764        struct patch_id_t data;
2765        char buffer[PATH_MAX * 4 + 20];
2766
2767        SHA1_Init(&ctx);
2768        memset(&data, 0, sizeof(struct patch_id_t));
2769        data.ctx = &ctx;
2770        data.xm.consume = patch_id_consume;
2771
2772        for (i = 0; i < q->nr; i++) {
2773                xpparam_t xpp;
2774                xdemitconf_t xecfg;
2775                xdemitcb_t ecb;
2776                mmfile_t mf1, mf2;
2777                struct diff_filepair *p = q->queue[i];
2778                int len1, len2;
2779
2780                memset(&xecfg, 0, sizeof(xecfg));
2781                if (p->status == 0)
2782                        return error("internal diff status error");
2783                if (p->status == DIFF_STATUS_UNKNOWN)
2784                        continue;
2785                if (diff_unmodified_pair(p))
2786                        continue;
2787                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2788                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2789                        continue;
2790                if (DIFF_PAIR_UNMERGED(p))
2791                        continue;
2792
2793                diff_fill_sha1_info(p->one);
2794                diff_fill_sha1_info(p->two);
2795                if (fill_mmfile(&mf1, p->one) < 0 ||
2796                                fill_mmfile(&mf2, p->two) < 0)
2797                        return error("unable to read files to diff");
2798
2799                len1 = remove_space(p->one->path, strlen(p->one->path));
2800                len2 = remove_space(p->two->path, strlen(p->two->path));
2801                if (p->one->mode == 0)
2802                        len1 = snprintf(buffer, sizeof(buffer),
2803                                        "diff--gita/%.*sb/%.*s"
2804                                        "newfilemode%06o"
2805                                        "---/dev/null"
2806                                        "+++b/%.*s",
2807                                        len1, p->one->path,
2808                                        len2, p->two->path,
2809                                        p->two->mode,
2810                                        len2, p->two->path);
2811                else if (p->two->mode == 0)
2812                        len1 = snprintf(buffer, sizeof(buffer),
2813                                        "diff--gita/%.*sb/%.*s"
2814                                        "deletedfilemode%06o"
2815                                        "---a/%.*s"
2816                                        "+++/dev/null",
2817                                        len1, p->one->path,
2818                                        len2, p->two->path,
2819                                        p->one->mode,
2820                                        len1, p->one->path);
2821                else
2822                        len1 = snprintf(buffer, sizeof(buffer),
2823                                        "diff--gita/%.*sb/%.*s"
2824                                        "---a/%.*s"
2825                                        "+++b/%.*s",
2826                                        len1, p->one->path,
2827                                        len2, p->two->path,
2828                                        len1, p->one->path,
2829                                        len2, p->two->path);
2830                SHA1_Update(&ctx, buffer, len1);
2831
2832                xpp.flags = XDF_NEED_MINIMAL;
2833                xecfg.ctxlen = 3;
2834                xecfg.flags = XDL_EMIT_FUNCNAMES;
2835                ecb.outf = xdiff_outf;
2836                ecb.priv = &data;
2837                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2838        }
2839
2840        SHA1_Final(sha1, &ctx);
2841        return 0;
2842}
2843
2844int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2845{
2846        struct diff_queue_struct *q = &diff_queued_diff;
2847        int i;
2848        int result = diff_get_patch_id(options, sha1);
2849
2850        for (i = 0; i < q->nr; i++)
2851                diff_free_filepair(q->queue[i]);
2852
2853        free(q->queue);
2854        q->queue = NULL;
2855        q->nr = q->alloc = 0;
2856
2857        return result;
2858}
2859
2860static int is_summary_empty(const struct diff_queue_struct *q)
2861{
2862        int i;
2863
2864        for (i = 0; i < q->nr; i++) {
2865                const struct diff_filepair *p = q->queue[i];
2866
2867                switch (p->status) {
2868                case DIFF_STATUS_DELETED:
2869                case DIFF_STATUS_ADDED:
2870                case DIFF_STATUS_COPIED:
2871                case DIFF_STATUS_RENAMED:
2872                        return 0;
2873                default:
2874                        if (p->score)
2875                                return 0;
2876                        if (p->one->mode && p->two->mode &&
2877                            p->one->mode != p->two->mode)
2878                                return 0;
2879                        break;
2880                }
2881        }
2882        return 1;
2883}
2884
2885void diff_flush(struct diff_options *options)
2886{
2887        struct diff_queue_struct *q = &diff_queued_diff;
2888        int i, output_format = options->output_format;
2889        int separator = 0;
2890
2891        /*
2892         * Order: raw, stat, summary, patch
2893         * or:    name/name-status/checkdiff (other bits clear)
2894         */
2895        if (!q->nr)
2896                goto free_queue;
2897
2898        if (output_format & (DIFF_FORMAT_RAW |
2899                             DIFF_FORMAT_NAME |
2900                             DIFF_FORMAT_NAME_STATUS |
2901                             DIFF_FORMAT_CHECKDIFF)) {
2902                for (i = 0; i < q->nr; i++) {
2903                        struct diff_filepair *p = q->queue[i];
2904                        if (check_pair_status(p))
2905                                flush_one_pair(p, options);
2906                }
2907                separator++;
2908        }
2909
2910        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2911                struct diffstat_t diffstat;
2912
2913                memset(&diffstat, 0, sizeof(struct diffstat_t));
2914                diffstat.xm.consume = diffstat_consume;
2915                for (i = 0; i < q->nr; i++) {
2916                        struct diff_filepair *p = q->queue[i];
2917                        if (check_pair_status(p))
2918                                diff_flush_stat(p, options, &diffstat);
2919                }
2920                if (output_format & DIFF_FORMAT_NUMSTAT)
2921                        show_numstat(&diffstat, options);
2922                if (output_format & DIFF_FORMAT_DIFFSTAT)
2923                        show_stats(&diffstat, options);
2924                if (output_format & DIFF_FORMAT_SHORTSTAT)
2925                        show_shortstats(&diffstat);
2926                free_diffstat_info(&diffstat);
2927                separator++;
2928        }
2929
2930        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2931                for (i = 0; i < q->nr; i++)
2932                        diff_summary(q->queue[i]);
2933                separator++;
2934        }
2935
2936        if (output_format & DIFF_FORMAT_PATCH) {
2937                if (separator) {
2938                        if (options->stat_sep) {
2939                                /* attach patch instead of inline */
2940                                fputs(options->stat_sep, stdout);
2941                        } else {
2942                                putchar(options->line_termination);
2943                        }
2944                }
2945
2946                for (i = 0; i < q->nr; i++) {
2947                        struct diff_filepair *p = q->queue[i];
2948                        if (check_pair_status(p))
2949                                diff_flush_patch(p, options);
2950                }
2951        }
2952
2953        if (output_format & DIFF_FORMAT_CALLBACK)
2954                options->format_callback(q, options, options->format_callback_data);
2955
2956        for (i = 0; i < q->nr; i++)
2957                diff_free_filepair(q->queue[i]);
2958free_queue:
2959        free(q->queue);
2960        q->queue = NULL;
2961        q->nr = q->alloc = 0;
2962}
2963
2964static void diffcore_apply_filter(const char *filter)
2965{
2966        int i;
2967        struct diff_queue_struct *q = &diff_queued_diff;
2968        struct diff_queue_struct outq;
2969        outq.queue = NULL;
2970        outq.nr = outq.alloc = 0;
2971
2972        if (!filter)
2973                return;
2974
2975        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2976                int found;
2977                for (i = found = 0; !found && i < q->nr; i++) {
2978                        struct diff_filepair *p = q->queue[i];
2979                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2980                             ((p->score &&
2981                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2982                              (!p->score &&
2983                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2984                            ((p->status != DIFF_STATUS_MODIFIED) &&
2985                             strchr(filter, p->status)))
2986                                found++;
2987                }
2988                if (found)
2989                        return;
2990
2991                /* otherwise we will clear the whole queue
2992                 * by copying the empty outq at the end of this
2993                 * function, but first clear the current entries
2994                 * in the queue.
2995                 */
2996                for (i = 0; i < q->nr; i++)
2997                        diff_free_filepair(q->queue[i]);
2998        }
2999        else {
3000                /* Only the matching ones */
3001                for (i = 0; i < q->nr; i++) {
3002                        struct diff_filepair *p = q->queue[i];
3003
3004                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3005                             ((p->score &&
3006                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3007                              (!p->score &&
3008                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3009                            ((p->status != DIFF_STATUS_MODIFIED) &&
3010                             strchr(filter, p->status)))
3011                                diff_q(&outq, p);
3012                        else
3013                                diff_free_filepair(p);
3014                }
3015        }
3016        free(q->queue);
3017        *q = outq;
3018}
3019
3020/* Check whether two filespecs with the same mode and size are identical */
3021static int diff_filespec_is_identical(struct diff_filespec *one,
3022                                      struct diff_filespec *two)
3023{
3024        if (S_ISGITLINK(one->mode)) {
3025                diff_fill_sha1_info(one);
3026                diff_fill_sha1_info(two);
3027                return !hashcmp(one->sha1, two->sha1);
3028        }
3029        if (diff_populate_filespec(one, 0))
3030                return 0;
3031        if (diff_populate_filespec(two, 0))
3032                return 0;
3033        return !memcmp(one->data, two->data, one->size);
3034}
3035
3036static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3037{
3038        int i;
3039        struct diff_queue_struct *q = &diff_queued_diff;
3040        struct diff_queue_struct outq;
3041        outq.queue = NULL;
3042        outq.nr = outq.alloc = 0;
3043
3044        for (i = 0; i < q->nr; i++) {
3045                struct diff_filepair *p = q->queue[i];
3046
3047                /*
3048                 * 1. Entries that come from stat info dirtyness
3049                 *    always have both sides (iow, not create/delete),
3050                 *    one side of the object name is unknown, with
3051                 *    the same mode and size.  Keep the ones that
3052                 *    do not match these criteria.  They have real
3053                 *    differences.
3054                 *
3055                 * 2. At this point, the file is known to be modified,
3056                 *    with the same mode and size, and the object
3057                 *    name of one side is unknown.  Need to inspect
3058                 *    the identical contents.
3059                 */
3060                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3061                    !DIFF_FILE_VALID(p->two) ||
3062                    (p->one->sha1_valid && p->two->sha1_valid) ||
3063                    (p->one->mode != p->two->mode) ||
3064                    diff_populate_filespec(p->one, 1) ||
3065                    diff_populate_filespec(p->two, 1) ||
3066                    (p->one->size != p->two->size) ||
3067                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3068                        diff_q(&outq, p);
3069                else {
3070                        /*
3071                         * The caller can subtract 1 from skip_stat_unmatch
3072                         * to determine how many paths were dirty only
3073                         * due to stat info mismatch.
3074                         */
3075                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3076                                diffopt->skip_stat_unmatch++;
3077                        diff_free_filepair(p);
3078                }
3079        }
3080        free(q->queue);
3081        *q = outq;
3082}
3083
3084void diffcore_std(struct diff_options *options)
3085{
3086        if (DIFF_OPT_TST(options, QUIET))
3087                return;
3088
3089        if (options->skip_stat_unmatch && !DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3090                diffcore_skip_stat_unmatch(options);
3091        if (options->break_opt != -1)
3092                diffcore_break(options->break_opt);
3093        if (options->detect_rename)
3094                diffcore_rename(options);
3095        if (options->break_opt != -1)
3096                diffcore_merge_broken();
3097        if (options->pickaxe)
3098                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3099        if (options->orderfile)
3100                diffcore_order(options->orderfile);
3101        diff_resolve_rename_copy();
3102        diffcore_apply_filter(options->filter);
3103
3104        if (diff_queued_diff.nr)
3105                DIFF_OPT_SET(options, HAS_CHANGES);
3106        else
3107                DIFF_OPT_CLR(options, HAS_CHANGES);
3108}
3109
3110int diff_result_code(struct diff_options *opt, int status)
3111{
3112        int result = 0;
3113        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3114            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3115                return status;
3116        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3117            DIFF_OPT_TST(opt, HAS_CHANGES))
3118                result |= 01;
3119        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3120            DIFF_OPT_TST(opt, CHECK_FAILED))
3121                result |= 02;
3122        return result;
3123}
3124
3125void diff_addremove(struct diff_options *options,
3126                    int addremove, unsigned mode,
3127                    const unsigned char *sha1,
3128                    const char *base, const char *path)
3129{
3130        char concatpath[PATH_MAX];
3131        struct diff_filespec *one, *two;
3132
3133        /* This may look odd, but it is a preparation for
3134         * feeding "there are unchanged files which should
3135         * not produce diffs, but when you are doing copy
3136         * detection you would need them, so here they are"
3137         * entries to the diff-core.  They will be prefixed
3138         * with something like '=' or '*' (I haven't decided
3139         * which but should not make any difference).
3140         * Feeding the same new and old to diff_change()
3141         * also has the same effect.
3142         * Before the final output happens, they are pruned after
3143         * merged into rename/copy pairs as appropriate.
3144         */
3145        if (DIFF_OPT_TST(options, REVERSE_DIFF))
3146                addremove = (addremove == '+' ? '-' :
3147                             addremove == '-' ? '+' : addremove);
3148
3149        if (!path) path = "";
3150        sprintf(concatpath, "%s%s", base, path);
3151        one = alloc_filespec(concatpath);
3152        two = alloc_filespec(concatpath);
3153
3154        if (addremove != '+')
3155                fill_filespec(one, sha1, mode);
3156        if (addremove != '-')
3157                fill_filespec(two, sha1, mode);
3158
3159        diff_queue(&diff_queued_diff, one, two);
3160        DIFF_OPT_SET(options, HAS_CHANGES);
3161}
3162
3163void diff_change(struct diff_options *options,
3164                 unsigned old_mode, unsigned new_mode,
3165                 const unsigned char *old_sha1,
3166                 const unsigned char *new_sha1,
3167                 const char *base, const char *path)
3168{
3169        char concatpath[PATH_MAX];
3170        struct diff_filespec *one, *two;
3171
3172        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3173                unsigned tmp;
3174                const unsigned char *tmp_c;
3175                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3176                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3177        }
3178        if (!path) path = "";
3179        sprintf(concatpath, "%s%s", base, path);
3180        one = alloc_filespec(concatpath);
3181        two = alloc_filespec(concatpath);
3182        fill_filespec(one, old_sha1, old_mode);
3183        fill_filespec(two, new_sha1, new_mode);
3184
3185        diff_queue(&diff_queued_diff, one, two);
3186        DIFF_OPT_SET(options, HAS_CHANGES);
3187}
3188
3189void diff_unmerge(struct diff_options *options,
3190                  const char *path,
3191                  unsigned mode, const unsigned char *sha1)
3192{
3193        struct diff_filespec *one, *two;
3194        one = alloc_filespec(path);
3195        two = alloc_filespec(path);
3196        fill_filespec(one, sha1, mode);
3197        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3198}