diff.con commit Unify whitespace checking (c1795bb)
   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;
  23int diff_auto_refresh_index = 1;
  24
  25static char diff_colors[][COLOR_MAXLEN] = {
  26        "\033[m",       /* reset */
  27        "",             /* PLAIN (normal) */
  28        "\033[1m",      /* METAINFO (bold) */
  29        "\033[36m",     /* FRAGINFO (cyan) */
  30        "\033[31m",     /* OLD (red) */
  31        "\033[32m",     /* NEW (green) */
  32        "\033[33m",     /* COMMIT (yellow) */
  33        "\033[41m",     /* WHITESPACE (red background) */
  34};
  35
  36static int parse_diff_color_slot(const char *var, int ofs)
  37{
  38        if (!strcasecmp(var+ofs, "plain"))
  39                return DIFF_PLAIN;
  40        if (!strcasecmp(var+ofs, "meta"))
  41                return DIFF_METAINFO;
  42        if (!strcasecmp(var+ofs, "frag"))
  43                return DIFF_FRAGINFO;
  44        if (!strcasecmp(var+ofs, "old"))
  45                return DIFF_FILE_OLD;
  46        if (!strcasecmp(var+ofs, "new"))
  47                return DIFF_FILE_NEW;
  48        if (!strcasecmp(var+ofs, "commit"))
  49                return DIFF_COMMIT;
  50        if (!strcasecmp(var+ofs, "whitespace"))
  51                return DIFF_WHITESPACE;
  52        die("bad config variable '%s'", var);
  53}
  54
  55static struct ll_diff_driver {
  56        const char *name;
  57        struct ll_diff_driver *next;
  58        char *cmd;
  59} *user_diff, **user_diff_tail;
  60
  61static void read_config_if_needed(void)
  62{
  63        if (!user_diff_tail) {
  64                user_diff_tail = &user_diff;
  65                git_config(git_diff_ui_config);
  66        }
  67}
  68
  69/*
  70 * Currently there is only "diff.<drivername>.command" variable;
  71 * because there are "diff.color.<slot>" variables, we are parsing
  72 * this in a bit convoluted way to allow low level diff driver
  73 * called "color".
  74 */
  75static int parse_lldiff_command(const char *var, const char *ep, const char *value)
  76{
  77        const char *name;
  78        int namelen;
  79        struct ll_diff_driver *drv;
  80
  81        name = var + 5;
  82        namelen = ep - name;
  83        for (drv = user_diff; drv; drv = drv->next)
  84                if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
  85                        break;
  86        if (!drv) {
  87                drv = xcalloc(1, sizeof(struct ll_diff_driver));
  88                drv->name = xmemdupz(name, namelen);
  89                if (!user_diff_tail)
  90                        user_diff_tail = &user_diff;
  91                *user_diff_tail = drv;
  92                user_diff_tail = &(drv->next);
  93        }
  94
  95        if (!value)
  96                return error("%s: lacks value", var);
  97        drv->cmd = strdup(value);
  98        return 0;
  99}
 100
 101/*
 102 * 'diff.<what>.funcname' attribute can be specified in the configuration
 103 * to define a customized regexp to find the beginning of a function to
 104 * be used for hunk header lines of "diff -p" style output.
 105 */
 106static struct funcname_pattern {
 107        char *name;
 108        char *pattern;
 109        struct funcname_pattern *next;
 110} *funcname_pattern_list;
 111
 112static int parse_funcname_pattern(const char *var, const char *ep, const char *value)
 113{
 114        const char *name;
 115        int namelen;
 116        struct funcname_pattern *pp;
 117
 118        name = var + 5; /* "diff." */
 119        namelen = ep - name;
 120
 121        for (pp = funcname_pattern_list; pp; pp = pp->next)
 122                if (!strncmp(pp->name, name, namelen) && !pp->name[namelen])
 123                        break;
 124        if (!pp) {
 125                pp = xcalloc(1, sizeof(*pp));
 126                pp->name = xmemdupz(name, namelen);
 127                pp->next = funcname_pattern_list;
 128                funcname_pattern_list = pp;
 129        }
 130        if (pp->pattern)
 131                free(pp->pattern);
 132        pp->pattern = xstrdup(value);
 133        return 0;
 134}
 135
 136/*
 137 * These are to give UI layer defaults.
 138 * The core-level commands such as git-diff-files should
 139 * never be affected by the setting of diff.renames
 140 * the user happens to have in the configuration file.
 141 */
 142int git_diff_ui_config(const char *var, const char *value)
 143{
 144        if (!strcmp(var, "diff.renamelimit")) {
 145                diff_rename_limit_default = git_config_int(var, value);
 146                return 0;
 147        }
 148        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 149                diff_use_color_default = git_config_colorbool(var, value, -1);
 150                return 0;
 151        }
 152        if (!strcmp(var, "diff.renames")) {
 153                if (!value)
 154                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 155                else if (!strcasecmp(value, "copies") ||
 156                         !strcasecmp(value, "copy"))
 157                        diff_detect_rename_default = DIFF_DETECT_COPY;
 158                else if (git_config_bool(var,value))
 159                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 160                return 0;
 161        }
 162        if (!strcmp(var, "diff.autorefreshindex")) {
 163                diff_auto_refresh_index = git_config_bool(var, value);
 164                return 0;
 165        }
 166        if (!prefixcmp(var, "diff.")) {
 167                const char *ep = strrchr(var, '.');
 168
 169                if (ep != var + 4) {
 170                        if (!strcmp(ep, ".command"))
 171                                return parse_lldiff_command(var, ep, value);
 172                        if (!strcmp(ep, ".funcname"))
 173                                return parse_funcname_pattern(var, ep, value);
 174                }
 175        }
 176        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 177                int slot = parse_diff_color_slot(var, 11);
 178                color_parse(value, var, diff_colors[slot]);
 179                return 0;
 180        }
 181
 182        return git_default_config(var, value);
 183}
 184
 185static char *quote_two(const char *one, const char *two)
 186{
 187        int need_one = quote_c_style(one, NULL, NULL, 1);
 188        int need_two = quote_c_style(two, NULL, NULL, 1);
 189        struct strbuf res;
 190
 191        strbuf_init(&res, 0);
 192        if (need_one + need_two) {
 193                strbuf_addch(&res, '"');
 194                quote_c_style(one, &res, NULL, 1);
 195                quote_c_style(two, &res, NULL, 1);
 196                strbuf_addch(&res, '"');
 197        } else {
 198                strbuf_addstr(&res, one);
 199                strbuf_addstr(&res, two);
 200        }
 201        return strbuf_detach(&res, NULL);
 202}
 203
 204static const char *external_diff(void)
 205{
 206        static const char *external_diff_cmd = NULL;
 207        static int done_preparing = 0;
 208
 209        if (done_preparing)
 210                return external_diff_cmd;
 211        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 212        done_preparing = 1;
 213        return external_diff_cmd;
 214}
 215
 216static struct diff_tempfile {
 217        const char *name; /* filename external diff should read from */
 218        char hex[41];
 219        char mode[10];
 220        char tmp_path[PATH_MAX];
 221} diff_temp[2];
 222
 223static int count_lines(const char *data, int size)
 224{
 225        int count, ch, completely_empty = 1, nl_just_seen = 0;
 226        count = 0;
 227        while (0 < size--) {
 228                ch = *data++;
 229                if (ch == '\n') {
 230                        count++;
 231                        nl_just_seen = 1;
 232                        completely_empty = 0;
 233                }
 234                else {
 235                        nl_just_seen = 0;
 236                        completely_empty = 0;
 237                }
 238        }
 239        if (completely_empty)
 240                return 0;
 241        if (!nl_just_seen)
 242                count++; /* no trailing newline */
 243        return count;
 244}
 245
 246static void print_line_count(int count)
 247{
 248        switch (count) {
 249        case 0:
 250                printf("0,0");
 251                break;
 252        case 1:
 253                printf("1");
 254                break;
 255        default:
 256                printf("1,%d", count);
 257                break;
 258        }
 259}
 260
 261static void copy_file(int prefix, const char *data, int size,
 262                const char *set, const char *reset)
 263{
 264        int ch, nl_just_seen = 1;
 265        while (0 < size--) {
 266                ch = *data++;
 267                if (nl_just_seen) {
 268                        fputs(set, stdout);
 269                        putchar(prefix);
 270                }
 271                if (ch == '\n') {
 272                        nl_just_seen = 1;
 273                        fputs(reset, stdout);
 274                } else
 275                        nl_just_seen = 0;
 276                putchar(ch);
 277        }
 278        if (!nl_just_seen)
 279                printf("%s\n\\ No newline at end of file\n", reset);
 280}
 281
 282static void emit_rewrite_diff(const char *name_a,
 283                              const char *name_b,
 284                              struct diff_filespec *one,
 285                              struct diff_filespec *two,
 286                              int color_diff)
 287{
 288        int lc_a, lc_b;
 289        const char *name_a_tab, *name_b_tab;
 290        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 291        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 292        const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
 293        const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
 294        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 295
 296        name_a += (*name_a == '/');
 297        name_b += (*name_b == '/');
 298        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 299        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 300
 301        diff_populate_filespec(one, 0);
 302        diff_populate_filespec(two, 0);
 303        lc_a = count_lines(one->data, one->size);
 304        lc_b = count_lines(two->data, two->size);
 305        printf("%s--- a/%s%s%s\n%s+++ b/%s%s%s\n%s@@ -",
 306               metainfo, name_a, name_a_tab, reset,
 307               metainfo, name_b, name_b_tab, reset, fraginfo);
 308        print_line_count(lc_a);
 309        printf(" +");
 310        print_line_count(lc_b);
 311        printf(" @@%s\n", reset);
 312        if (lc_a)
 313                copy_file('-', one->data, one->size, old, reset);
 314        if (lc_b)
 315                copy_file('+', two->data, two->size, new, reset);
 316}
 317
 318static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 319{
 320        if (!DIFF_FILE_VALID(one)) {
 321                mf->ptr = (char *)""; /* does not matter */
 322                mf->size = 0;
 323                return 0;
 324        }
 325        else if (diff_populate_filespec(one, 0))
 326                return -1;
 327        mf->ptr = one->data;
 328        mf->size = one->size;
 329        return 0;
 330}
 331
 332struct diff_words_buffer {
 333        mmfile_t text;
 334        long alloc;
 335        long current; /* output pointer */
 336        int suppressed_newline;
 337};
 338
 339static void diff_words_append(char *line, unsigned long len,
 340                struct diff_words_buffer *buffer)
 341{
 342        if (buffer->text.size + len > buffer->alloc) {
 343                buffer->alloc = (buffer->text.size + len) * 3 / 2;
 344                buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
 345        }
 346        line++;
 347        len--;
 348        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 349        buffer->text.size += len;
 350}
 351
 352struct diff_words_data {
 353        struct xdiff_emit_state xm;
 354        struct diff_words_buffer minus, plus;
 355};
 356
 357static void print_word(struct diff_words_buffer *buffer, int len, int color,
 358                int suppress_newline)
 359{
 360        const char *ptr;
 361        int eol = 0;
 362
 363        if (len == 0)
 364                return;
 365
 366        ptr  = buffer->text.ptr + buffer->current;
 367        buffer->current += len;
 368
 369        if (ptr[len - 1] == '\n') {
 370                eol = 1;
 371                len--;
 372        }
 373
 374        fputs(diff_get_color(1, color), stdout);
 375        fwrite(ptr, len, 1, stdout);
 376        fputs(diff_get_color(1, DIFF_RESET), stdout);
 377
 378        if (eol) {
 379                if (suppress_newline)
 380                        buffer->suppressed_newline = 1;
 381                else
 382                        putchar('\n');
 383        }
 384}
 385
 386static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 387{
 388        struct diff_words_data *diff_words = priv;
 389
 390        if (diff_words->minus.suppressed_newline) {
 391                if (line[0] != '+')
 392                        putchar('\n');
 393                diff_words->minus.suppressed_newline = 0;
 394        }
 395
 396        len--;
 397        switch (line[0]) {
 398                case '-':
 399                        print_word(&diff_words->minus, len, DIFF_FILE_OLD, 1);
 400                        break;
 401                case '+':
 402                        print_word(&diff_words->plus, len, DIFF_FILE_NEW, 0);
 403                        break;
 404                case ' ':
 405                        print_word(&diff_words->plus, len, DIFF_PLAIN, 0);
 406                        diff_words->minus.current += len;
 407                        break;
 408        }
 409}
 410
 411/* this executes the word diff on the accumulated buffers */
 412static void diff_words_show(struct diff_words_data *diff_words)
 413{
 414        xpparam_t xpp;
 415        xdemitconf_t xecfg;
 416        xdemitcb_t ecb;
 417        mmfile_t minus, plus;
 418        int i;
 419
 420        memset(&xecfg, 0, sizeof(xecfg));
 421        minus.size = diff_words->minus.text.size;
 422        minus.ptr = xmalloc(minus.size);
 423        memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
 424        for (i = 0; i < minus.size; i++)
 425                if (isspace(minus.ptr[i]))
 426                        minus.ptr[i] = '\n';
 427        diff_words->minus.current = 0;
 428
 429        plus.size = diff_words->plus.text.size;
 430        plus.ptr = xmalloc(plus.size);
 431        memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
 432        for (i = 0; i < plus.size; i++)
 433                if (isspace(plus.ptr[i]))
 434                        plus.ptr[i] = '\n';
 435        diff_words->plus.current = 0;
 436
 437        xpp.flags = XDF_NEED_MINIMAL;
 438        xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
 439        ecb.outf = xdiff_outf;
 440        ecb.priv = diff_words;
 441        diff_words->xm.consume = fn_out_diff_words_aux;
 442        xdi_diff(&minus, &plus, &xpp, &xecfg, &ecb);
 443
 444        free(minus.ptr);
 445        free(plus.ptr);
 446        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 447
 448        if (diff_words->minus.suppressed_newline) {
 449                putchar('\n');
 450                diff_words->minus.suppressed_newline = 0;
 451        }
 452}
 453
 454struct emit_callback {
 455        struct xdiff_emit_state xm;
 456        int nparents, color_diff;
 457        unsigned ws_rule;
 458        const char **label_path;
 459        struct diff_words_data *diff_words;
 460        int *found_changesp;
 461};
 462
 463static void free_diff_words_data(struct emit_callback *ecbdata)
 464{
 465        if (ecbdata->diff_words) {
 466                /* flush buffers */
 467                if (ecbdata->diff_words->minus.text.size ||
 468                                ecbdata->diff_words->plus.text.size)
 469                        diff_words_show(ecbdata->diff_words);
 470
 471                if (ecbdata->diff_words->minus.text.ptr)
 472                        free (ecbdata->diff_words->minus.text.ptr);
 473                if (ecbdata->diff_words->plus.text.ptr)
 474                        free (ecbdata->diff_words->plus.text.ptr);
 475                free(ecbdata->diff_words);
 476                ecbdata->diff_words = NULL;
 477        }
 478}
 479
 480const char *diff_get_color(int diff_use_color, enum color_diff ix)
 481{
 482        if (diff_use_color)
 483                return diff_colors[ix];
 484        return "";
 485}
 486
 487static void emit_line(const char *set, const char *reset, const char *line, int len)
 488{
 489        fputs(set, stdout);
 490        fwrite(line, len, 1, stdout);
 491        fputs(reset, stdout);
 492}
 493
 494static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 495{
 496        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 497        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 498
 499        if (!*ws)
 500                emit_line(set, reset, line, len);
 501        else {
 502                /* Emit just the prefix, then the rest. */
 503                emit_line(set, reset, line, ecbdata->nparents);
 504                (void)check_and_emit_line(line + ecbdata->nparents,
 505                    len - ecbdata->nparents, ecbdata->ws_rule,
 506                    stdout, set, reset, ws);
 507        }
 508}
 509
 510static void fn_out_consume(void *priv, char *line, unsigned long len)
 511{
 512        int i;
 513        int color;
 514        struct emit_callback *ecbdata = priv;
 515        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 516        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 517
 518        *(ecbdata->found_changesp) = 1;
 519
 520        if (ecbdata->label_path[0]) {
 521                const char *name_a_tab, *name_b_tab;
 522
 523                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 524                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 525
 526                printf("%s--- %s%s%s\n",
 527                       set, ecbdata->label_path[0], reset, name_a_tab);
 528                printf("%s+++ %s%s%s\n",
 529                       set, ecbdata->label_path[1], reset, name_b_tab);
 530                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 531        }
 532
 533        /* This is not really necessary for now because
 534         * this codepath only deals with two-way diffs.
 535         */
 536        for (i = 0; i < len && line[i] == '@'; i++)
 537                ;
 538        if (2 <= i && i < len && line[i] == ' ') {
 539                ecbdata->nparents = i - 1;
 540                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 541                          reset, line, len);
 542                return;
 543        }
 544
 545        if (len < ecbdata->nparents) {
 546                set = reset;
 547                emit_line(reset, reset, line, len);
 548                return;
 549        }
 550
 551        color = DIFF_PLAIN;
 552        if (ecbdata->diff_words && ecbdata->nparents != 1)
 553                /* fall back to normal diff */
 554                free_diff_words_data(ecbdata);
 555        if (ecbdata->diff_words) {
 556                if (line[0] == '-') {
 557                        diff_words_append(line, len,
 558                                          &ecbdata->diff_words->minus);
 559                        return;
 560                } else if (line[0] == '+') {
 561                        diff_words_append(line, len,
 562                                          &ecbdata->diff_words->plus);
 563                        return;
 564                }
 565                if (ecbdata->diff_words->minus.text.size ||
 566                    ecbdata->diff_words->plus.text.size)
 567                        diff_words_show(ecbdata->diff_words);
 568                line++;
 569                len--;
 570                emit_line(set, reset, line, len);
 571                return;
 572        }
 573        for (i = 0; i < ecbdata->nparents && len; i++) {
 574                if (line[i] == '-')
 575                        color = DIFF_FILE_OLD;
 576                else if (line[i] == '+')
 577                        color = DIFF_FILE_NEW;
 578        }
 579
 580        if (color != DIFF_FILE_NEW) {
 581                emit_line(diff_get_color(ecbdata->color_diff, color),
 582                          reset, line, len);
 583                return;
 584        }
 585        emit_add_line(reset, ecbdata, line, len);
 586}
 587
 588static char *pprint_rename(const char *a, const char *b)
 589{
 590        const char *old = a;
 591        const char *new = b;
 592        struct strbuf name;
 593        int pfx_length, sfx_length;
 594        int len_a = strlen(a);
 595        int len_b = strlen(b);
 596        int a_midlen, b_midlen;
 597        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 598        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 599
 600        strbuf_init(&name, 0);
 601        if (qlen_a || qlen_b) {
 602                quote_c_style(a, &name, NULL, 0);
 603                strbuf_addstr(&name, " => ");
 604                quote_c_style(b, &name, NULL, 0);
 605                return strbuf_detach(&name, NULL);
 606        }
 607
 608        /* Find common prefix */
 609        pfx_length = 0;
 610        while (*old && *new && *old == *new) {
 611                if (*old == '/')
 612                        pfx_length = old - a + 1;
 613                old++;
 614                new++;
 615        }
 616
 617        /* Find common suffix */
 618        old = a + len_a;
 619        new = b + len_b;
 620        sfx_length = 0;
 621        while (a <= old && b <= new && *old == *new) {
 622                if (*old == '/')
 623                        sfx_length = len_a - (old - a);
 624                old--;
 625                new--;
 626        }
 627
 628        /*
 629         * pfx{mid-a => mid-b}sfx
 630         * {pfx-a => pfx-b}sfx
 631         * pfx{sfx-a => sfx-b}
 632         * name-a => name-b
 633         */
 634        a_midlen = len_a - pfx_length - sfx_length;
 635        b_midlen = len_b - pfx_length - sfx_length;
 636        if (a_midlen < 0)
 637                a_midlen = 0;
 638        if (b_midlen < 0)
 639                b_midlen = 0;
 640
 641        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 642        if (pfx_length + sfx_length) {
 643                strbuf_add(&name, a, pfx_length);
 644                strbuf_addch(&name, '{');
 645        }
 646        strbuf_add(&name, a + pfx_length, a_midlen);
 647        strbuf_addstr(&name, " => ");
 648        strbuf_add(&name, b + pfx_length, b_midlen);
 649        if (pfx_length + sfx_length) {
 650                strbuf_addch(&name, '}');
 651                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 652        }
 653        return strbuf_detach(&name, NULL);
 654}
 655
 656struct diffstat_t {
 657        struct xdiff_emit_state xm;
 658
 659        int nr;
 660        int alloc;
 661        struct diffstat_file {
 662                char *from_name;
 663                char *name;
 664                char *print_name;
 665                unsigned is_unmerged:1;
 666                unsigned is_binary:1;
 667                unsigned is_renamed:1;
 668                unsigned int added, deleted;
 669        } **files;
 670};
 671
 672static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 673                                          const char *name_a,
 674                                          const char *name_b)
 675{
 676        struct diffstat_file *x;
 677        x = xcalloc(sizeof (*x), 1);
 678        if (diffstat->nr == diffstat->alloc) {
 679                diffstat->alloc = alloc_nr(diffstat->alloc);
 680                diffstat->files = xrealloc(diffstat->files,
 681                                diffstat->alloc * sizeof(x));
 682        }
 683        diffstat->files[diffstat->nr++] = x;
 684        if (name_b) {
 685                x->from_name = xstrdup(name_a);
 686                x->name = xstrdup(name_b);
 687                x->is_renamed = 1;
 688        }
 689        else {
 690                x->from_name = NULL;
 691                x->name = xstrdup(name_a);
 692        }
 693        return x;
 694}
 695
 696static void diffstat_consume(void *priv, char *line, unsigned long len)
 697{
 698        struct diffstat_t *diffstat = priv;
 699        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 700
 701        if (line[0] == '+')
 702                x->added++;
 703        else if (line[0] == '-')
 704                x->deleted++;
 705}
 706
 707const char mime_boundary_leader[] = "------------";
 708
 709static int scale_linear(int it, int width, int max_change)
 710{
 711        /*
 712         * make sure that at least one '-' is printed if there were deletions,
 713         * and likewise for '+'.
 714         */
 715        if (max_change < 2)
 716                return it;
 717        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 718}
 719
 720static void show_name(const char *prefix, const char *name, int len,
 721                      const char *reset, const char *set)
 722{
 723        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 724}
 725
 726static void show_graph(char ch, int cnt, const char *set, const char *reset)
 727{
 728        if (cnt <= 0)
 729                return;
 730        printf("%s", set);
 731        while (cnt--)
 732                putchar(ch);
 733        printf("%s", reset);
 734}
 735
 736static void fill_print_name(struct diffstat_file *file)
 737{
 738        char *pname;
 739
 740        if (file->print_name)
 741                return;
 742
 743        if (!file->is_renamed) {
 744                struct strbuf buf;
 745                strbuf_init(&buf, 0);
 746                if (quote_c_style(file->name, &buf, NULL, 0)) {
 747                        pname = strbuf_detach(&buf, NULL);
 748                } else {
 749                        pname = file->name;
 750                        strbuf_release(&buf);
 751                }
 752        } else {
 753                pname = pprint_rename(file->from_name, file->name);
 754        }
 755        file->print_name = pname;
 756}
 757
 758static void show_stats(struct diffstat_t* data, struct diff_options *options)
 759{
 760        int i, len, add, del, total, adds = 0, dels = 0;
 761        int max_change = 0, max_len = 0;
 762        int total_files = data->nr;
 763        int width, name_width;
 764        const char *reset, *set, *add_c, *del_c;
 765
 766        if (data->nr == 0)
 767                return;
 768
 769        width = options->stat_width ? options->stat_width : 80;
 770        name_width = options->stat_name_width ? options->stat_name_width : 50;
 771
 772        /* Sanity: give at least 5 columns to the graph,
 773         * but leave at least 10 columns for the name.
 774         */
 775        if (width < name_width + 15) {
 776                if (name_width <= 25)
 777                        width = name_width + 15;
 778                else
 779                        name_width = width - 15;
 780        }
 781
 782        /* Find the longest filename and max number of changes */
 783        reset = diff_get_color_opt(options, DIFF_RESET);
 784        set   = diff_get_color_opt(options, DIFF_PLAIN);
 785        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
 786        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
 787
 788        for (i = 0; i < data->nr; i++) {
 789                struct diffstat_file *file = data->files[i];
 790                int change = file->added + file->deleted;
 791                fill_print_name(file);
 792                len = strlen(file->print_name);
 793                if (max_len < len)
 794                        max_len = len;
 795
 796                if (file->is_binary || file->is_unmerged)
 797                        continue;
 798                if (max_change < change)
 799                        max_change = change;
 800        }
 801
 802        /* Compute the width of the graph part;
 803         * 10 is for one blank at the beginning of the line plus
 804         * " | count " between the name and the graph.
 805         *
 806         * From here on, name_width is the width of the name area,
 807         * and width is the width of the graph area.
 808         */
 809        name_width = (name_width < max_len) ? name_width : max_len;
 810        if (width < (name_width + 10) + max_change)
 811                width = width - (name_width + 10);
 812        else
 813                width = max_change;
 814
 815        for (i = 0; i < data->nr; i++) {
 816                const char *prefix = "";
 817                char *name = data->files[i]->print_name;
 818                int added = data->files[i]->added;
 819                int deleted = data->files[i]->deleted;
 820                int name_len;
 821
 822                /*
 823                 * "scale" the filename
 824                 */
 825                len = name_width;
 826                name_len = strlen(name);
 827                if (name_width < name_len) {
 828                        char *slash;
 829                        prefix = "...";
 830                        len -= 3;
 831                        name += name_len - len;
 832                        slash = strchr(name, '/');
 833                        if (slash)
 834                                name = slash;
 835                }
 836
 837                if (data->files[i]->is_binary) {
 838                        show_name(prefix, name, len, reset, set);
 839                        printf("  Bin ");
 840                        printf("%s%d%s", del_c, deleted, reset);
 841                        printf(" -> ");
 842                        printf("%s%d%s", add_c, added, reset);
 843                        printf(" bytes");
 844                        printf("\n");
 845                        continue;
 846                }
 847                else if (data->files[i]->is_unmerged) {
 848                        show_name(prefix, name, len, reset, set);
 849                        printf("  Unmerged\n");
 850                        continue;
 851                }
 852                else if (!data->files[i]->is_renamed &&
 853                         (added + deleted == 0)) {
 854                        total_files--;
 855                        continue;
 856                }
 857
 858                /*
 859                 * scale the add/delete
 860                 */
 861                add = added;
 862                del = deleted;
 863                total = add + del;
 864                adds += add;
 865                dels += del;
 866
 867                if (width <= max_change) {
 868                        add = scale_linear(add, width, max_change);
 869                        del = scale_linear(del, width, max_change);
 870                        total = add + del;
 871                }
 872                show_name(prefix, name, len, reset, set);
 873                printf("%5d ", added + deleted);
 874                show_graph('+', add, add_c, reset);
 875                show_graph('-', del, del_c, reset);
 876                putchar('\n');
 877        }
 878        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 879               set, total_files, adds, dels, reset);
 880}
 881
 882static void show_shortstats(struct diffstat_t* data)
 883{
 884        int i, adds = 0, dels = 0, total_files = data->nr;
 885
 886        if (data->nr == 0)
 887                return;
 888
 889        for (i = 0; i < data->nr; i++) {
 890                if (!data->files[i]->is_binary &&
 891                    !data->files[i]->is_unmerged) {
 892                        int added = data->files[i]->added;
 893                        int deleted= data->files[i]->deleted;
 894                        if (!data->files[i]->is_renamed &&
 895                            (added + deleted == 0)) {
 896                                total_files--;
 897                        } else {
 898                                adds += added;
 899                                dels += deleted;
 900                        }
 901                }
 902        }
 903        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 904               total_files, adds, dels);
 905}
 906
 907static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 908{
 909        int i;
 910
 911        if (data->nr == 0)
 912                return;
 913
 914        for (i = 0; i < data->nr; i++) {
 915                struct diffstat_file *file = data->files[i];
 916
 917                if (file->is_binary)
 918                        printf("-\t-\t");
 919                else
 920                        printf("%d\t%d\t", file->added, file->deleted);
 921                if (options->line_termination) {
 922                        fill_print_name(file);
 923                        if (!file->is_renamed)
 924                                write_name_quoted(file->name, stdout,
 925                                                  options->line_termination);
 926                        else {
 927                                fputs(file->print_name, stdout);
 928                                putchar(options->line_termination);
 929                        }
 930                } else {
 931                        if (file->is_renamed) {
 932                                putchar('\0');
 933                                write_name_quoted(file->from_name, stdout, '\0');
 934                        }
 935                        write_name_quoted(file->name, stdout, '\0');
 936                }
 937        }
 938}
 939
 940static void free_diffstat_info(struct diffstat_t *diffstat)
 941{
 942        int i;
 943        for (i = 0; i < diffstat->nr; i++) {
 944                struct diffstat_file *f = diffstat->files[i];
 945                if (f->name != f->print_name)
 946                        free(f->print_name);
 947                free(f->name);
 948                free(f->from_name);
 949                free(f);
 950        }
 951        free(diffstat->files);
 952}
 953
 954struct checkdiff_t {
 955        struct xdiff_emit_state xm;
 956        const char *filename;
 957        int lineno, color_diff;
 958        unsigned ws_rule;
 959        unsigned status;
 960};
 961
 962static void checkdiff_consume(void *priv, char *line, unsigned long len)
 963{
 964        struct checkdiff_t *data = priv;
 965        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
 966        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
 967        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
 968        char *err;
 969
 970        if (line[0] == '+') {
 971                data->status = check_and_emit_line(line + 1, len - 1,
 972                    data->ws_rule, NULL, NULL, NULL, NULL);
 973                if (!data->status)
 974                        return;
 975                err = whitespace_error_string(data->status);
 976                printf("%s:%d: %s%s:%s ", data->filename, data->lineno,
 977                    ws, err, reset);
 978                free(err);
 979                emit_line(set, reset, line, 1);
 980                (void)check_and_emit_line(line + 1, len - 1, data->ws_rule,
 981                    stdout, set, reset, ws);
 982                data->lineno++;
 983        } else if (line[0] == ' ')
 984                data->lineno++;
 985        else if (line[0] == '@') {
 986                char *plus = strchr(line, '+');
 987                if (plus)
 988                        data->lineno = strtol(plus, NULL, 10);
 989                else
 990                        die("invalid diff");
 991        }
 992}
 993
 994static unsigned char *deflate_it(char *data,
 995                                 unsigned long size,
 996                                 unsigned long *result_size)
 997{
 998        int bound;
 999        unsigned char *deflated;
1000        z_stream stream;
1001
1002        memset(&stream, 0, sizeof(stream));
1003        deflateInit(&stream, zlib_compression_level);
1004        bound = deflateBound(&stream, size);
1005        deflated = xmalloc(bound);
1006        stream.next_out = deflated;
1007        stream.avail_out = bound;
1008
1009        stream.next_in = (unsigned char *)data;
1010        stream.avail_in = size;
1011        while (deflate(&stream, Z_FINISH) == Z_OK)
1012                ; /* nothing */
1013        deflateEnd(&stream);
1014        *result_size = stream.total_out;
1015        return deflated;
1016}
1017
1018static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1019{
1020        void *cp;
1021        void *delta;
1022        void *deflated;
1023        void *data;
1024        unsigned long orig_size;
1025        unsigned long delta_size;
1026        unsigned long deflate_size;
1027        unsigned long data_size;
1028
1029        /* We could do deflated delta, or we could do just deflated two,
1030         * whichever is smaller.
1031         */
1032        delta = NULL;
1033        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1034        if (one->size && two->size) {
1035                delta = diff_delta(one->ptr, one->size,
1036                                   two->ptr, two->size,
1037                                   &delta_size, deflate_size);
1038                if (delta) {
1039                        void *to_free = delta;
1040                        orig_size = delta_size;
1041                        delta = deflate_it(delta, delta_size, &delta_size);
1042                        free(to_free);
1043                }
1044        }
1045
1046        if (delta && delta_size < deflate_size) {
1047                printf("delta %lu\n", orig_size);
1048                free(deflated);
1049                data = delta;
1050                data_size = delta_size;
1051        }
1052        else {
1053                printf("literal %lu\n", two->size);
1054                free(delta);
1055                data = deflated;
1056                data_size = deflate_size;
1057        }
1058
1059        /* emit data encoded in base85 */
1060        cp = data;
1061        while (data_size) {
1062                int bytes = (52 < data_size) ? 52 : data_size;
1063                char line[70];
1064                data_size -= bytes;
1065                if (bytes <= 26)
1066                        line[0] = bytes + 'A' - 1;
1067                else
1068                        line[0] = bytes - 26 + 'a' - 1;
1069                encode_85(line + 1, cp, bytes);
1070                cp = (char *) cp + bytes;
1071                puts(line);
1072        }
1073        printf("\n");
1074        free(data);
1075}
1076
1077static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1078{
1079        printf("GIT binary patch\n");
1080        emit_binary_diff_body(one, two);
1081        emit_binary_diff_body(two, one);
1082}
1083
1084static void setup_diff_attr_check(struct git_attr_check *check)
1085{
1086        static struct git_attr *attr_diff;
1087
1088        if (!attr_diff) {
1089                attr_diff = git_attr("diff", 4);
1090        }
1091        check[0].attr = attr_diff;
1092}
1093
1094static void diff_filespec_check_attr(struct diff_filespec *one)
1095{
1096        struct git_attr_check attr_diff_check;
1097        int check_from_data = 0;
1098
1099        if (one->checked_attr)
1100                return;
1101
1102        setup_diff_attr_check(&attr_diff_check);
1103        one->is_binary = 0;
1104        one->funcname_pattern_ident = NULL;
1105
1106        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1107                const char *value;
1108
1109                /* binaryness */
1110                value = attr_diff_check.value;
1111                if (ATTR_TRUE(value))
1112                        ;
1113                else if (ATTR_FALSE(value))
1114                        one->is_binary = 1;
1115                else
1116                        check_from_data = 1;
1117
1118                /* funcname pattern ident */
1119                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1120                        ;
1121                else
1122                        one->funcname_pattern_ident = value;
1123        }
1124
1125        if (check_from_data) {
1126                if (!one->data && DIFF_FILE_VALID(one))
1127                        diff_populate_filespec(one, 0);
1128
1129                if (one->data)
1130                        one->is_binary = buffer_is_binary(one->data, one->size);
1131        }
1132}
1133
1134int diff_filespec_is_binary(struct diff_filespec *one)
1135{
1136        diff_filespec_check_attr(one);
1137        return one->is_binary;
1138}
1139
1140static const char *funcname_pattern(const char *ident)
1141{
1142        struct funcname_pattern *pp;
1143
1144        read_config_if_needed();
1145        for (pp = funcname_pattern_list; pp; pp = pp->next)
1146                if (!strcmp(ident, pp->name))
1147                        return pp->pattern;
1148        return NULL;
1149}
1150
1151static struct builtin_funcname_pattern {
1152        const char *name;
1153        const char *pattern;
1154} builtin_funcname_pattern[] = {
1155        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1156                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1157                        "^[     ]*\\(\\([       ]*"
1158                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1159                        "[      ]*([^;]*$\\)" },
1160        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1161};
1162
1163static const char *diff_funcname_pattern(struct diff_filespec *one)
1164{
1165        const char *ident, *pattern;
1166        int i;
1167
1168        diff_filespec_check_attr(one);
1169        ident = one->funcname_pattern_ident;
1170
1171        if (!ident)
1172                /*
1173                 * If the config file has "funcname.default" defined, that
1174                 * regexp is used; otherwise NULL is returned and xemit uses
1175                 * the built-in default.
1176                 */
1177                return funcname_pattern("default");
1178
1179        /* Look up custom "funcname.$ident" regexp from config. */
1180        pattern = funcname_pattern(ident);
1181        if (pattern)
1182                return pattern;
1183
1184        /*
1185         * And define built-in fallback patterns here.  Note that
1186         * these can be overriden by the user's config settings.
1187         */
1188        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1189                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1190                        return builtin_funcname_pattern[i].pattern;
1191
1192        return NULL;
1193}
1194
1195static void builtin_diff(const char *name_a,
1196                         const char *name_b,
1197                         struct diff_filespec *one,
1198                         struct diff_filespec *two,
1199                         const char *xfrm_msg,
1200                         struct diff_options *o,
1201                         int complete_rewrite)
1202{
1203        mmfile_t mf1, mf2;
1204        const char *lbl[2];
1205        char *a_one, *b_two;
1206        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1207        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1208
1209        a_one = quote_two("a/", name_a + (*name_a == '/'));
1210        b_two = quote_two("b/", name_b + (*name_b == '/'));
1211        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1212        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1213        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1214        if (lbl[0][0] == '/') {
1215                /* /dev/null */
1216                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1217                if (xfrm_msg && xfrm_msg[0])
1218                        printf("%s%s%s\n", set, xfrm_msg, reset);
1219        }
1220        else if (lbl[1][0] == '/') {
1221                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1222                if (xfrm_msg && xfrm_msg[0])
1223                        printf("%s%s%s\n", set, xfrm_msg, reset);
1224        }
1225        else {
1226                if (one->mode != two->mode) {
1227                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1228                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1229                }
1230                if (xfrm_msg && xfrm_msg[0])
1231                        printf("%s%s%s\n", set, xfrm_msg, reset);
1232                /*
1233                 * we do not run diff between different kind
1234                 * of objects.
1235                 */
1236                if ((one->mode ^ two->mode) & S_IFMT)
1237                        goto free_ab_and_return;
1238                if (complete_rewrite) {
1239                        emit_rewrite_diff(name_a, name_b, one, two,
1240                                        DIFF_OPT_TST(o, COLOR_DIFF));
1241                        o->found_changes = 1;
1242                        goto free_ab_and_return;
1243                }
1244        }
1245
1246        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1247                die("unable to read files to diff");
1248
1249        if (!DIFF_OPT_TST(o, TEXT) &&
1250            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1251                /* Quite common confusing case */
1252                if (mf1.size == mf2.size &&
1253                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1254                        goto free_ab_and_return;
1255                if (DIFF_OPT_TST(o, BINARY))
1256                        emit_binary_diff(&mf1, &mf2);
1257                else
1258                        printf("Binary files %s and %s differ\n",
1259                               lbl[0], lbl[1]);
1260                o->found_changes = 1;
1261        }
1262        else {
1263                /* Crazy xdl interfaces.. */
1264                const char *diffopts = getenv("GIT_DIFF_OPTS");
1265                xpparam_t xpp;
1266                xdemitconf_t xecfg;
1267                xdemitcb_t ecb;
1268                struct emit_callback ecbdata;
1269                const char *funcname_pattern;
1270
1271                funcname_pattern = diff_funcname_pattern(one);
1272                if (!funcname_pattern)
1273                        funcname_pattern = diff_funcname_pattern(two);
1274
1275                memset(&xecfg, 0, sizeof(xecfg));
1276                memset(&ecbdata, 0, sizeof(ecbdata));
1277                ecbdata.label_path = lbl;
1278                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1279                ecbdata.found_changesp = &o->found_changes;
1280                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1281                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1282                xecfg.ctxlen = o->context;
1283                xecfg.flags = XDL_EMIT_FUNCNAMES;
1284                if (funcname_pattern)
1285                        xdiff_set_find_func(&xecfg, funcname_pattern);
1286                if (!diffopts)
1287                        ;
1288                else if (!prefixcmp(diffopts, "--unified="))
1289                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1290                else if (!prefixcmp(diffopts, "-u"))
1291                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1292                ecb.outf = xdiff_outf;
1293                ecb.priv = &ecbdata;
1294                ecbdata.xm.consume = fn_out_consume;
1295                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1296                        ecbdata.diff_words =
1297                                xcalloc(1, sizeof(struct diff_words_data));
1298                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1299                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1300                        free_diff_words_data(&ecbdata);
1301        }
1302
1303 free_ab_and_return:
1304        diff_free_filespec_data(one);
1305        diff_free_filespec_data(two);
1306        free(a_one);
1307        free(b_two);
1308        return;
1309}
1310
1311static void builtin_diffstat(const char *name_a, const char *name_b,
1312                             struct diff_filespec *one,
1313                             struct diff_filespec *two,
1314                             struct diffstat_t *diffstat,
1315                             struct diff_options *o,
1316                             int complete_rewrite)
1317{
1318        mmfile_t mf1, mf2;
1319        struct diffstat_file *data;
1320
1321        data = diffstat_add(diffstat, name_a, name_b);
1322
1323        if (!one || !two) {
1324                data->is_unmerged = 1;
1325                return;
1326        }
1327        if (complete_rewrite) {
1328                diff_populate_filespec(one, 0);
1329                diff_populate_filespec(two, 0);
1330                data->deleted = count_lines(one->data, one->size);
1331                data->added = count_lines(two->data, two->size);
1332                goto free_and_return;
1333        }
1334        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1335                die("unable to read files to diff");
1336
1337        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1338                data->is_binary = 1;
1339                data->added = mf2.size;
1340                data->deleted = mf1.size;
1341        } else {
1342                /* Crazy xdl interfaces.. */
1343                xpparam_t xpp;
1344                xdemitconf_t xecfg;
1345                xdemitcb_t ecb;
1346
1347                memset(&xecfg, 0, sizeof(xecfg));
1348                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1349                ecb.outf = xdiff_outf;
1350                ecb.priv = diffstat;
1351                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1352        }
1353
1354 free_and_return:
1355        diff_free_filespec_data(one);
1356        diff_free_filespec_data(two);
1357}
1358
1359static void builtin_checkdiff(const char *name_a, const char *name_b,
1360                             struct diff_filespec *one,
1361                             struct diff_filespec *two, struct diff_options *o)
1362{
1363        mmfile_t mf1, mf2;
1364        struct checkdiff_t data;
1365
1366        if (!two)
1367                return;
1368
1369        memset(&data, 0, sizeof(data));
1370        data.xm.consume = checkdiff_consume;
1371        data.filename = name_b ? name_b : name_a;
1372        data.lineno = 0;
1373        data.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1374        data.ws_rule = whitespace_rule(data.filename);
1375
1376        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1377                die("unable to read files to diff");
1378
1379        if (diff_filespec_is_binary(two))
1380                goto free_and_return;
1381        else {
1382                /* Crazy xdl interfaces.. */
1383                xpparam_t xpp;
1384                xdemitconf_t xecfg;
1385                xdemitcb_t ecb;
1386
1387                memset(&xecfg, 0, sizeof(xecfg));
1388                xpp.flags = XDF_NEED_MINIMAL;
1389                ecb.outf = xdiff_outf;
1390                ecb.priv = &data;
1391                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1392        }
1393 free_and_return:
1394        diff_free_filespec_data(one);
1395        diff_free_filespec_data(two);
1396        if (data.status)
1397                DIFF_OPT_SET(o, CHECK_FAILED);
1398}
1399
1400struct diff_filespec *alloc_filespec(const char *path)
1401{
1402        int namelen = strlen(path);
1403        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1404
1405        memset(spec, 0, sizeof(*spec));
1406        spec->path = (char *)(spec + 1);
1407        memcpy(spec->path, path, namelen+1);
1408        spec->count = 1;
1409        return spec;
1410}
1411
1412void free_filespec(struct diff_filespec *spec)
1413{
1414        if (!--spec->count) {
1415                diff_free_filespec_data(spec);
1416                free(spec);
1417        }
1418}
1419
1420void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1421                   unsigned short mode)
1422{
1423        if (mode) {
1424                spec->mode = canon_mode(mode);
1425                hashcpy(spec->sha1, sha1);
1426                spec->sha1_valid = !is_null_sha1(sha1);
1427        }
1428}
1429
1430/*
1431 * Given a name and sha1 pair, if the index tells us the file in
1432 * the work tree has that object contents, return true, so that
1433 * prepare_temp_file() does not have to inflate and extract.
1434 */
1435static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1436{
1437        struct cache_entry *ce;
1438        struct stat st;
1439        int pos, len;
1440
1441        /* We do not read the cache ourselves here, because the
1442         * benchmark with my previous version that always reads cache
1443         * shows that it makes things worse for diff-tree comparing
1444         * two linux-2.6 kernel trees in an already checked out work
1445         * tree.  This is because most diff-tree comparisons deal with
1446         * only a small number of files, while reading the cache is
1447         * expensive for a large project, and its cost outweighs the
1448         * savings we get by not inflating the object to a temporary
1449         * file.  Practically, this code only helps when we are used
1450         * by diff-cache --cached, which does read the cache before
1451         * calling us.
1452         */
1453        if (!active_cache)
1454                return 0;
1455
1456        /* We want to avoid the working directory if our caller
1457         * doesn't need the data in a normal file, this system
1458         * is rather slow with its stat/open/mmap/close syscalls,
1459         * and the object is contained in a pack file.  The pack
1460         * is probably already open and will be faster to obtain
1461         * the data through than the working directory.  Loose
1462         * objects however would tend to be slower as they need
1463         * to be individually opened and inflated.
1464         */
1465        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1466                return 0;
1467
1468        len = strlen(name);
1469        pos = cache_name_pos(name, len);
1470        if (pos < 0)
1471                return 0;
1472        ce = active_cache[pos];
1473        if ((lstat(name, &st) < 0) ||
1474            !S_ISREG(st.st_mode) || /* careful! */
1475            ce_match_stat(ce, &st, 0) ||
1476            hashcmp(sha1, ce->sha1))
1477                return 0;
1478        /* we return 1 only when we can stat, it is a regular file,
1479         * stat information matches, and sha1 recorded in the cache
1480         * matches.  I.e. we know the file in the work tree really is
1481         * the same as the <name, sha1> pair.
1482         */
1483        return 1;
1484}
1485
1486static int populate_from_stdin(struct diff_filespec *s)
1487{
1488        struct strbuf buf;
1489        size_t size = 0;
1490
1491        strbuf_init(&buf, 0);
1492        if (strbuf_read(&buf, 0, 0) < 0)
1493                return error("error while reading from stdin %s",
1494                                     strerror(errno));
1495
1496        s->should_munmap = 0;
1497        s->data = strbuf_detach(&buf, &size);
1498        s->size = size;
1499        s->should_free = 1;
1500        return 0;
1501}
1502
1503static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1504{
1505        int len;
1506        char *data = xmalloc(100);
1507        len = snprintf(data, 100,
1508                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1509        s->data = data;
1510        s->size = len;
1511        s->should_free = 1;
1512        if (size_only) {
1513                s->data = NULL;
1514                free(data);
1515        }
1516        return 0;
1517}
1518
1519/*
1520 * While doing rename detection and pickaxe operation, we may need to
1521 * grab the data for the blob (or file) for our own in-core comparison.
1522 * diff_filespec has data and size fields for this purpose.
1523 */
1524int diff_populate_filespec(struct diff_filespec *s, int size_only)
1525{
1526        int err = 0;
1527        if (!DIFF_FILE_VALID(s))
1528                die("internal error: asking to populate invalid file.");
1529        if (S_ISDIR(s->mode))
1530                return -1;
1531
1532        if (s->data)
1533                return 0;
1534
1535        if (size_only && 0 < s->size)
1536                return 0;
1537
1538        if (S_ISGITLINK(s->mode))
1539                return diff_populate_gitlink(s, size_only);
1540
1541        if (!s->sha1_valid ||
1542            reuse_worktree_file(s->path, s->sha1, 0)) {
1543                struct strbuf buf;
1544                struct stat st;
1545                int fd;
1546
1547                if (!strcmp(s->path, "-"))
1548                        return populate_from_stdin(s);
1549
1550                if (lstat(s->path, &st) < 0) {
1551                        if (errno == ENOENT) {
1552                        err_empty:
1553                                err = -1;
1554                        empty:
1555                                s->data = (char *)"";
1556                                s->size = 0;
1557                                return err;
1558                        }
1559                }
1560                s->size = xsize_t(st.st_size);
1561                if (!s->size)
1562                        goto empty;
1563                if (size_only)
1564                        return 0;
1565                if (S_ISLNK(st.st_mode)) {
1566                        int ret;
1567                        s->data = xmalloc(s->size);
1568                        s->should_free = 1;
1569                        ret = readlink(s->path, s->data, s->size);
1570                        if (ret < 0) {
1571                                free(s->data);
1572                                goto err_empty;
1573                        }
1574                        return 0;
1575                }
1576                fd = open(s->path, O_RDONLY);
1577                if (fd < 0)
1578                        goto err_empty;
1579                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1580                close(fd);
1581                s->should_munmap = 1;
1582
1583                /*
1584                 * Convert from working tree format to canonical git format
1585                 */
1586                strbuf_init(&buf, 0);
1587                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1588                        size_t size = 0;
1589                        munmap(s->data, s->size);
1590                        s->should_munmap = 0;
1591                        s->data = strbuf_detach(&buf, &size);
1592                        s->size = size;
1593                        s->should_free = 1;
1594                }
1595        }
1596        else {
1597                enum object_type type;
1598                if (size_only)
1599                        type = sha1_object_info(s->sha1, &s->size);
1600                else {
1601                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1602                        s->should_free = 1;
1603                }
1604        }
1605        return 0;
1606}
1607
1608void diff_free_filespec_blob(struct diff_filespec *s)
1609{
1610        if (s->should_free)
1611                free(s->data);
1612        else if (s->should_munmap)
1613                munmap(s->data, s->size);
1614
1615        if (s->should_free || s->should_munmap) {
1616                s->should_free = s->should_munmap = 0;
1617                s->data = NULL;
1618        }
1619}
1620
1621void diff_free_filespec_data(struct diff_filespec *s)
1622{
1623        diff_free_filespec_blob(s);
1624        free(s->cnt_data);
1625        s->cnt_data = NULL;
1626}
1627
1628static void prep_temp_blob(struct diff_tempfile *temp,
1629                           void *blob,
1630                           unsigned long size,
1631                           const unsigned char *sha1,
1632                           int mode)
1633{
1634        int fd;
1635
1636        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1637        if (fd < 0)
1638                die("unable to create temp-file: %s", strerror(errno));
1639        if (write_in_full(fd, blob, size) != size)
1640                die("unable to write temp-file");
1641        close(fd);
1642        temp->name = temp->tmp_path;
1643        strcpy(temp->hex, sha1_to_hex(sha1));
1644        temp->hex[40] = 0;
1645        sprintf(temp->mode, "%06o", mode);
1646}
1647
1648static void prepare_temp_file(const char *name,
1649                              struct diff_tempfile *temp,
1650                              struct diff_filespec *one)
1651{
1652        if (!DIFF_FILE_VALID(one)) {
1653        not_a_valid_file:
1654                /* A '-' entry produces this for file-2, and
1655                 * a '+' entry produces this for file-1.
1656                 */
1657                temp->name = "/dev/null";
1658                strcpy(temp->hex, ".");
1659                strcpy(temp->mode, ".");
1660                return;
1661        }
1662
1663        if (!one->sha1_valid ||
1664            reuse_worktree_file(name, one->sha1, 1)) {
1665                struct stat st;
1666                if (lstat(name, &st) < 0) {
1667                        if (errno == ENOENT)
1668                                goto not_a_valid_file;
1669                        die("stat(%s): %s", name, strerror(errno));
1670                }
1671                if (S_ISLNK(st.st_mode)) {
1672                        int ret;
1673                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1674                        size_t sz = xsize_t(st.st_size);
1675                        if (sizeof(buf) <= st.st_size)
1676                                die("symlink too long: %s", name);
1677                        ret = readlink(name, buf, sz);
1678                        if (ret < 0)
1679                                die("readlink(%s)", name);
1680                        prep_temp_blob(temp, buf, sz,
1681                                       (one->sha1_valid ?
1682                                        one->sha1 : null_sha1),
1683                                       (one->sha1_valid ?
1684                                        one->mode : S_IFLNK));
1685                }
1686                else {
1687                        /* we can borrow from the file in the work tree */
1688                        temp->name = name;
1689                        if (!one->sha1_valid)
1690                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1691                        else
1692                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1693                        /* Even though we may sometimes borrow the
1694                         * contents from the work tree, we always want
1695                         * one->mode.  mode is trustworthy even when
1696                         * !(one->sha1_valid), as long as
1697                         * DIFF_FILE_VALID(one).
1698                         */
1699                        sprintf(temp->mode, "%06o", one->mode);
1700                }
1701                return;
1702        }
1703        else {
1704                if (diff_populate_filespec(one, 0))
1705                        die("cannot read data blob for %s", one->path);
1706                prep_temp_blob(temp, one->data, one->size,
1707                               one->sha1, one->mode);
1708        }
1709}
1710
1711static void remove_tempfile(void)
1712{
1713        int i;
1714
1715        for (i = 0; i < 2; i++)
1716                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1717                        unlink(diff_temp[i].name);
1718                        diff_temp[i].name = NULL;
1719                }
1720}
1721
1722static void remove_tempfile_on_signal(int signo)
1723{
1724        remove_tempfile();
1725        signal(SIGINT, SIG_DFL);
1726        raise(signo);
1727}
1728
1729/* An external diff command takes:
1730 *
1731 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1732 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1733 *
1734 */
1735static void run_external_diff(const char *pgm,
1736                              const char *name,
1737                              const char *other,
1738                              struct diff_filespec *one,
1739                              struct diff_filespec *two,
1740                              const char *xfrm_msg,
1741                              int complete_rewrite)
1742{
1743        const char *spawn_arg[10];
1744        struct diff_tempfile *temp = diff_temp;
1745        int retval;
1746        static int atexit_asked = 0;
1747        const char *othername;
1748        const char **arg = &spawn_arg[0];
1749
1750        othername = (other? other : name);
1751        if (one && two) {
1752                prepare_temp_file(name, &temp[0], one);
1753                prepare_temp_file(othername, &temp[1], two);
1754                if (! atexit_asked &&
1755                    (temp[0].name == temp[0].tmp_path ||
1756                     temp[1].name == temp[1].tmp_path)) {
1757                        atexit_asked = 1;
1758                        atexit(remove_tempfile);
1759                }
1760                signal(SIGINT, remove_tempfile_on_signal);
1761        }
1762
1763        if (one && two) {
1764                *arg++ = pgm;
1765                *arg++ = name;
1766                *arg++ = temp[0].name;
1767                *arg++ = temp[0].hex;
1768                *arg++ = temp[0].mode;
1769                *arg++ = temp[1].name;
1770                *arg++ = temp[1].hex;
1771                *arg++ = temp[1].mode;
1772                if (other) {
1773                        *arg++ = other;
1774                        *arg++ = xfrm_msg;
1775                }
1776        } else {
1777                *arg++ = pgm;
1778                *arg++ = name;
1779        }
1780        *arg = NULL;
1781        fflush(NULL);
1782        retval = run_command_v_opt(spawn_arg, 0);
1783        remove_tempfile();
1784        if (retval) {
1785                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1786                exit(1);
1787        }
1788}
1789
1790static const char *external_diff_attr(const char *name)
1791{
1792        struct git_attr_check attr_diff_check;
1793
1794        setup_diff_attr_check(&attr_diff_check);
1795        if (!git_checkattr(name, 1, &attr_diff_check)) {
1796                const char *value = attr_diff_check.value;
1797                if (!ATTR_TRUE(value) &&
1798                    !ATTR_FALSE(value) &&
1799                    !ATTR_UNSET(value)) {
1800                        struct ll_diff_driver *drv;
1801
1802                        read_config_if_needed();
1803                        for (drv = user_diff; drv; drv = drv->next)
1804                                if (!strcmp(drv->name, value))
1805                                        return drv->cmd;
1806                }
1807        }
1808        return NULL;
1809}
1810
1811static void run_diff_cmd(const char *pgm,
1812                         const char *name,
1813                         const char *other,
1814                         struct diff_filespec *one,
1815                         struct diff_filespec *two,
1816                         const char *xfrm_msg,
1817                         struct diff_options *o,
1818                         int complete_rewrite)
1819{
1820        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
1821                pgm = NULL;
1822        else {
1823                const char *cmd = external_diff_attr(name);
1824                if (cmd)
1825                        pgm = cmd;
1826        }
1827
1828        if (pgm) {
1829                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1830                                  complete_rewrite);
1831                return;
1832        }
1833        if (one && two)
1834                builtin_diff(name, other ? other : name,
1835                             one, two, xfrm_msg, o, complete_rewrite);
1836        else
1837                printf("* Unmerged path %s\n", name);
1838}
1839
1840static void diff_fill_sha1_info(struct diff_filespec *one)
1841{
1842        if (DIFF_FILE_VALID(one)) {
1843                if (!one->sha1_valid) {
1844                        struct stat st;
1845                        if (!strcmp(one->path, "-")) {
1846                                hashcpy(one->sha1, null_sha1);
1847                                return;
1848                        }
1849                        if (lstat(one->path, &st) < 0)
1850                                die("stat %s", one->path);
1851                        if (index_path(one->sha1, one->path, &st, 0))
1852                                die("cannot hash %s\n", one->path);
1853                }
1854        }
1855        else
1856                hashclr(one->sha1);
1857}
1858
1859static int similarity_index(struct diff_filepair *p)
1860{
1861        return p->score * 100 / MAX_SCORE;
1862}
1863
1864static void run_diff(struct diff_filepair *p, struct diff_options *o)
1865{
1866        const char *pgm = external_diff();
1867        struct strbuf msg;
1868        char *xfrm_msg;
1869        struct diff_filespec *one = p->one;
1870        struct diff_filespec *two = p->two;
1871        const char *name;
1872        const char *other;
1873        int complete_rewrite = 0;
1874
1875
1876        if (DIFF_PAIR_UNMERGED(p)) {
1877                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1878                return;
1879        }
1880
1881        name  = p->one->path;
1882        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1883        diff_fill_sha1_info(one);
1884        diff_fill_sha1_info(two);
1885
1886        strbuf_init(&msg, PATH_MAX * 2 + 300);
1887        switch (p->status) {
1888        case DIFF_STATUS_COPIED:
1889                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1890                strbuf_addstr(&msg, "\ncopy from ");
1891                quote_c_style(name, &msg, NULL, 0);
1892                strbuf_addstr(&msg, "\ncopy to ");
1893                quote_c_style(other, &msg, NULL, 0);
1894                strbuf_addch(&msg, '\n');
1895                break;
1896        case DIFF_STATUS_RENAMED:
1897                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1898                strbuf_addstr(&msg, "\nrename from ");
1899                quote_c_style(name, &msg, NULL, 0);
1900                strbuf_addstr(&msg, "\nrename to ");
1901                quote_c_style(other, &msg, NULL, 0);
1902                strbuf_addch(&msg, '\n');
1903                break;
1904        case DIFF_STATUS_MODIFIED:
1905                if (p->score) {
1906                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
1907                                        similarity_index(p));
1908                        complete_rewrite = 1;
1909                        break;
1910                }
1911                /* fallthru */
1912        default:
1913                /* nothing */
1914                ;
1915        }
1916
1917        if (hashcmp(one->sha1, two->sha1)) {
1918                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
1919
1920                if (DIFF_OPT_TST(o, BINARY)) {
1921                        mmfile_t mf;
1922                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
1923                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
1924                                abbrev = 40;
1925                }
1926                strbuf_addf(&msg, "index %.*s..%.*s",
1927                                abbrev, sha1_to_hex(one->sha1),
1928                                abbrev, sha1_to_hex(two->sha1));
1929                if (one->mode == two->mode)
1930                        strbuf_addf(&msg, " %06o", one->mode);
1931                strbuf_addch(&msg, '\n');
1932        }
1933
1934        if (msg.len)
1935                strbuf_setlen(&msg, msg.len - 1);
1936        xfrm_msg = msg.len ? msg.buf : NULL;
1937
1938        if (!pgm &&
1939            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1940            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1941                /* a filepair that changes between file and symlink
1942                 * needs to be split into deletion and creation.
1943                 */
1944                struct diff_filespec *null = alloc_filespec(two->path);
1945                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1946                free(null);
1947                null = alloc_filespec(one->path);
1948                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1949                free(null);
1950        }
1951        else
1952                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
1953                             complete_rewrite);
1954
1955        strbuf_release(&msg);
1956}
1957
1958static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
1959                         struct diffstat_t *diffstat)
1960{
1961        const char *name;
1962        const char *other;
1963        int complete_rewrite = 0;
1964
1965        if (DIFF_PAIR_UNMERGED(p)) {
1966                /* unmerged */
1967                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
1968                return;
1969        }
1970
1971        name = p->one->path;
1972        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1973
1974        diff_fill_sha1_info(p->one);
1975        diff_fill_sha1_info(p->two);
1976
1977        if (p->status == DIFF_STATUS_MODIFIED && p->score)
1978                complete_rewrite = 1;
1979        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
1980}
1981
1982static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
1983{
1984        const char *name;
1985        const char *other;
1986
1987        if (DIFF_PAIR_UNMERGED(p)) {
1988                /* unmerged */
1989                return;
1990        }
1991
1992        name = p->one->path;
1993        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1994
1995        diff_fill_sha1_info(p->one);
1996        diff_fill_sha1_info(p->two);
1997
1998        builtin_checkdiff(name, other, p->one, p->two, o);
1999}
2000
2001void diff_setup(struct diff_options *options)
2002{
2003        memset(options, 0, sizeof(*options));
2004        options->line_termination = '\n';
2005        options->break_opt = -1;
2006        options->rename_limit = -1;
2007        options->context = 3;
2008        options->msg_sep = "";
2009
2010        options->change = diff_change;
2011        options->add_remove = diff_addremove;
2012        if (diff_use_color_default)
2013                DIFF_OPT_SET(options, COLOR_DIFF);
2014        else
2015                DIFF_OPT_CLR(options, COLOR_DIFF);
2016        options->detect_rename = diff_detect_rename_default;
2017}
2018
2019int diff_setup_done(struct diff_options *options)
2020{
2021        int count = 0;
2022
2023        if (options->output_format & DIFF_FORMAT_NAME)
2024                count++;
2025        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2026                count++;
2027        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2028                count++;
2029        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2030                count++;
2031        if (count > 1)
2032                die("--name-only, --name-status, --check and -s are mutually exclusive");
2033
2034        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2035                options->detect_rename = DIFF_DETECT_COPY;
2036
2037        if (options->output_format & (DIFF_FORMAT_NAME |
2038                                      DIFF_FORMAT_NAME_STATUS |
2039                                      DIFF_FORMAT_CHECKDIFF |
2040                                      DIFF_FORMAT_NO_OUTPUT))
2041                options->output_format &= ~(DIFF_FORMAT_RAW |
2042                                            DIFF_FORMAT_NUMSTAT |
2043                                            DIFF_FORMAT_DIFFSTAT |
2044                                            DIFF_FORMAT_SHORTSTAT |
2045                                            DIFF_FORMAT_SUMMARY |
2046                                            DIFF_FORMAT_PATCH);
2047
2048        /*
2049         * These cases always need recursive; we do not drop caller-supplied
2050         * recursive bits for other formats here.
2051         */
2052        if (options->output_format & (DIFF_FORMAT_PATCH |
2053                                      DIFF_FORMAT_NUMSTAT |
2054                                      DIFF_FORMAT_DIFFSTAT |
2055                                      DIFF_FORMAT_SHORTSTAT |
2056                                      DIFF_FORMAT_SUMMARY |
2057                                      DIFF_FORMAT_CHECKDIFF))
2058                DIFF_OPT_SET(options, RECURSIVE);
2059        /*
2060         * Also pickaxe would not work very well if you do not say recursive
2061         */
2062        if (options->pickaxe)
2063                DIFF_OPT_SET(options, RECURSIVE);
2064
2065        if (options->detect_rename && options->rename_limit < 0)
2066                options->rename_limit = diff_rename_limit_default;
2067        if (options->setup & DIFF_SETUP_USE_CACHE) {
2068                if (!active_cache)
2069                        /* read-cache does not die even when it fails
2070                         * so it is safe for us to do this here.  Also
2071                         * it does not smudge active_cache or active_nr
2072                         * when it fails, so we do not have to worry about
2073                         * cleaning it up ourselves either.
2074                         */
2075                        read_cache();
2076        }
2077        if (options->abbrev <= 0 || 40 < options->abbrev)
2078                options->abbrev = 40; /* full */
2079
2080        /*
2081         * It does not make sense to show the first hit we happened
2082         * to have found.  It does not make sense not to return with
2083         * exit code in such a case either.
2084         */
2085        if (DIFF_OPT_TST(options, QUIET)) {
2086                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2087                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2088        }
2089
2090        /*
2091         * If we postprocess in diffcore, we cannot simply return
2092         * upon the first hit.  We need to run diff as usual.
2093         */
2094        if (options->pickaxe || options->filter)
2095                DIFF_OPT_CLR(options, QUIET);
2096
2097        return 0;
2098}
2099
2100static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2101{
2102        char c, *eq;
2103        int len;
2104
2105        if (*arg != '-')
2106                return 0;
2107        c = *++arg;
2108        if (!c)
2109                return 0;
2110        if (c == arg_short) {
2111                c = *++arg;
2112                if (!c)
2113                        return 1;
2114                if (val && isdigit(c)) {
2115                        char *end;
2116                        int n = strtoul(arg, &end, 10);
2117                        if (*end)
2118                                return 0;
2119                        *val = n;
2120                        return 1;
2121                }
2122                return 0;
2123        }
2124        if (c != '-')
2125                return 0;
2126        arg++;
2127        eq = strchr(arg, '=');
2128        if (eq)
2129                len = eq - arg;
2130        else
2131                len = strlen(arg);
2132        if (!len || strncmp(arg, arg_long, len))
2133                return 0;
2134        if (eq) {
2135                int n;
2136                char *end;
2137                if (!isdigit(*++eq))
2138                        return 0;
2139                n = strtoul(eq, &end, 10);
2140                if (*end)
2141                        return 0;
2142                *val = n;
2143        }
2144        return 1;
2145}
2146
2147static int diff_scoreopt_parse(const char *opt);
2148
2149int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2150{
2151        const char *arg = av[0];
2152
2153        /* Output format options */
2154        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2155                options->output_format |= DIFF_FORMAT_PATCH;
2156        else if (opt_arg(arg, 'U', "unified", &options->context))
2157                options->output_format |= DIFF_FORMAT_PATCH;
2158        else if (!strcmp(arg, "--raw"))
2159                options->output_format |= DIFF_FORMAT_RAW;
2160        else if (!strcmp(arg, "--patch-with-raw"))
2161                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2162        else if (!strcmp(arg, "--numstat"))
2163                options->output_format |= DIFF_FORMAT_NUMSTAT;
2164        else if (!strcmp(arg, "--shortstat"))
2165                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2166        else if (!strcmp(arg, "--check"))
2167                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2168        else if (!strcmp(arg, "--summary"))
2169                options->output_format |= DIFF_FORMAT_SUMMARY;
2170        else if (!strcmp(arg, "--patch-with-stat"))
2171                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2172        else if (!strcmp(arg, "--name-only"))
2173                options->output_format |= DIFF_FORMAT_NAME;
2174        else if (!strcmp(arg, "--name-status"))
2175                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2176        else if (!strcmp(arg, "-s"))
2177                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2178        else if (!prefixcmp(arg, "--stat")) {
2179                char *end;
2180                int width = options->stat_width;
2181                int name_width = options->stat_name_width;
2182                arg += 6;
2183                end = (char *)arg;
2184
2185                switch (*arg) {
2186                case '-':
2187                        if (!prefixcmp(arg, "-width="))
2188                                width = strtoul(arg + 7, &end, 10);
2189                        else if (!prefixcmp(arg, "-name-width="))
2190                                name_width = strtoul(arg + 12, &end, 10);
2191                        break;
2192                case '=':
2193                        width = strtoul(arg+1, &end, 10);
2194                        if (*end == ',')
2195                                name_width = strtoul(end+1, &end, 10);
2196                }
2197
2198                /* Important! This checks all the error cases! */
2199                if (*end)
2200                        return 0;
2201                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2202                options->stat_name_width = name_width;
2203                options->stat_width = width;
2204        }
2205
2206        /* renames options */
2207        else if (!prefixcmp(arg, "-B")) {
2208                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2209                        return -1;
2210        }
2211        else if (!prefixcmp(arg, "-M")) {
2212                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2213                        return -1;
2214                options->detect_rename = DIFF_DETECT_RENAME;
2215        }
2216        else if (!prefixcmp(arg, "-C")) {
2217                if (options->detect_rename == DIFF_DETECT_COPY)
2218                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2219                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2220                        return -1;
2221                options->detect_rename = DIFF_DETECT_COPY;
2222        }
2223        else if (!strcmp(arg, "--no-renames"))
2224                options->detect_rename = 0;
2225
2226        /* xdiff options */
2227        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2228                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2229        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2230                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2231        else if (!strcmp(arg, "--ignore-space-at-eol"))
2232                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2233
2234        /* flags options */
2235        else if (!strcmp(arg, "--binary")) {
2236                options->output_format |= DIFF_FORMAT_PATCH;
2237                DIFF_OPT_SET(options, BINARY);
2238        }
2239        else if (!strcmp(arg, "--full-index"))
2240                DIFF_OPT_SET(options, FULL_INDEX);
2241        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2242                DIFF_OPT_SET(options, TEXT);
2243        else if (!strcmp(arg, "-R"))
2244                DIFF_OPT_SET(options, REVERSE_DIFF);
2245        else if (!strcmp(arg, "--find-copies-harder"))
2246                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2247        else if (!strcmp(arg, "--follow"))
2248                DIFF_OPT_SET(options, FOLLOW_RENAMES);
2249        else if (!strcmp(arg, "--color"))
2250                DIFF_OPT_SET(options, COLOR_DIFF);
2251        else if (!strcmp(arg, "--no-color"))
2252                DIFF_OPT_CLR(options, COLOR_DIFF);
2253        else if (!strcmp(arg, "--color-words"))
2254                options->flags |= DIFF_OPT_COLOR_DIFF | DIFF_OPT_COLOR_DIFF_WORDS;
2255        else if (!strcmp(arg, "--exit-code"))
2256                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2257        else if (!strcmp(arg, "--quiet"))
2258                DIFF_OPT_SET(options, QUIET);
2259        else if (!strcmp(arg, "--ext-diff"))
2260                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2261        else if (!strcmp(arg, "--no-ext-diff"))
2262                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2263
2264        /* misc options */
2265        else if (!strcmp(arg, "-z"))
2266                options->line_termination = 0;
2267        else if (!prefixcmp(arg, "-l"))
2268                options->rename_limit = strtoul(arg+2, NULL, 10);
2269        else if (!prefixcmp(arg, "-S"))
2270                options->pickaxe = arg + 2;
2271        else if (!strcmp(arg, "--pickaxe-all"))
2272                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2273        else if (!strcmp(arg, "--pickaxe-regex"))
2274                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2275        else if (!prefixcmp(arg, "-O"))
2276                options->orderfile = arg + 2;
2277        else if (!prefixcmp(arg, "--diff-filter="))
2278                options->filter = arg + 14;
2279        else if (!strcmp(arg, "--abbrev"))
2280                options->abbrev = DEFAULT_ABBREV;
2281        else if (!prefixcmp(arg, "--abbrev=")) {
2282                options->abbrev = strtoul(arg + 9, NULL, 10);
2283                if (options->abbrev < MINIMUM_ABBREV)
2284                        options->abbrev = MINIMUM_ABBREV;
2285                else if (40 < options->abbrev)
2286                        options->abbrev = 40;
2287        }
2288        else
2289                return 0;
2290        return 1;
2291}
2292
2293static int parse_num(const char **cp_p)
2294{
2295        unsigned long num, scale;
2296        int ch, dot;
2297        const char *cp = *cp_p;
2298
2299        num = 0;
2300        scale = 1;
2301        dot = 0;
2302        for(;;) {
2303                ch = *cp;
2304                if ( !dot && ch == '.' ) {
2305                        scale = 1;
2306                        dot = 1;
2307                } else if ( ch == '%' ) {
2308                        scale = dot ? scale*100 : 100;
2309                        cp++;   /* % is always at the end */
2310                        break;
2311                } else if ( ch >= '0' && ch <= '9' ) {
2312                        if ( scale < 100000 ) {
2313                                scale *= 10;
2314                                num = (num*10) + (ch-'0');
2315                        }
2316                } else {
2317                        break;
2318                }
2319                cp++;
2320        }
2321        *cp_p = cp;
2322
2323        /* user says num divided by scale and we say internally that
2324         * is MAX_SCORE * num / scale.
2325         */
2326        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2327}
2328
2329static int diff_scoreopt_parse(const char *opt)
2330{
2331        int opt1, opt2, cmd;
2332
2333        if (*opt++ != '-')
2334                return -1;
2335        cmd = *opt++;
2336        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2337                return -1; /* that is not a -M, -C nor -B option */
2338
2339        opt1 = parse_num(&opt);
2340        if (cmd != 'B')
2341                opt2 = 0;
2342        else {
2343                if (*opt == 0)
2344                        opt2 = 0;
2345                else if (*opt != '/')
2346                        return -1; /* we expect -B80/99 or -B80 */
2347                else {
2348                        opt++;
2349                        opt2 = parse_num(&opt);
2350                }
2351        }
2352        if (*opt != 0)
2353                return -1;
2354        return opt1 | (opt2 << 16);
2355}
2356
2357struct diff_queue_struct diff_queued_diff;
2358
2359void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2360{
2361        if (queue->alloc <= queue->nr) {
2362                queue->alloc = alloc_nr(queue->alloc);
2363                queue->queue = xrealloc(queue->queue,
2364                                        sizeof(dp) * queue->alloc);
2365        }
2366        queue->queue[queue->nr++] = dp;
2367}
2368
2369struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2370                                 struct diff_filespec *one,
2371                                 struct diff_filespec *two)
2372{
2373        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2374        dp->one = one;
2375        dp->two = two;
2376        if (queue)
2377                diff_q(queue, dp);
2378        return dp;
2379}
2380
2381void diff_free_filepair(struct diff_filepair *p)
2382{
2383        free_filespec(p->one);
2384        free_filespec(p->two);
2385        free(p);
2386}
2387
2388/* This is different from find_unique_abbrev() in that
2389 * it stuffs the result with dots for alignment.
2390 */
2391const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2392{
2393        int abblen;
2394        const char *abbrev;
2395        if (len == 40)
2396                return sha1_to_hex(sha1);
2397
2398        abbrev = find_unique_abbrev(sha1, len);
2399        if (!abbrev)
2400                return sha1_to_hex(sha1);
2401        abblen = strlen(abbrev);
2402        if (abblen < 37) {
2403                static char hex[41];
2404                if (len < abblen && abblen <= len + 2)
2405                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2406                else
2407                        sprintf(hex, "%s...", abbrev);
2408                return hex;
2409        }
2410        return sha1_to_hex(sha1);
2411}
2412
2413static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2414{
2415        int line_termination = opt->line_termination;
2416        int inter_name_termination = line_termination ? '\t' : '\0';
2417
2418        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2419                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2420                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2421                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2422        }
2423        if (p->score) {
2424                printf("%c%03d%c", p->status, similarity_index(p),
2425                           inter_name_termination);
2426        } else {
2427                printf("%c%c", p->status, inter_name_termination);
2428        }
2429
2430        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2431                write_name_quoted(p->one->path, stdout, inter_name_termination);
2432                write_name_quoted(p->two->path, stdout, line_termination);
2433        } else {
2434                const char *path = p->one->mode ? p->one->path : p->two->path;
2435                write_name_quoted(path, stdout, line_termination);
2436        }
2437}
2438
2439int diff_unmodified_pair(struct diff_filepair *p)
2440{
2441        /* This function is written stricter than necessary to support
2442         * the currently implemented transformers, but the idea is to
2443         * let transformers to produce diff_filepairs any way they want,
2444         * and filter and clean them up here before producing the output.
2445         */
2446        struct diff_filespec *one = p->one, *two = p->two;
2447
2448        if (DIFF_PAIR_UNMERGED(p))
2449                return 0; /* unmerged is interesting */
2450
2451        /* deletion, addition, mode or type change
2452         * and rename are all interesting.
2453         */
2454        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2455            DIFF_PAIR_MODE_CHANGED(p) ||
2456            strcmp(one->path, two->path))
2457                return 0;
2458
2459        /* both are valid and point at the same path.  that is, we are
2460         * dealing with a change.
2461         */
2462        if (one->sha1_valid && two->sha1_valid &&
2463            !hashcmp(one->sha1, two->sha1))
2464                return 1; /* no change */
2465        if (!one->sha1_valid && !two->sha1_valid)
2466                return 1; /* both look at the same file on the filesystem. */
2467        return 0;
2468}
2469
2470static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2471{
2472        if (diff_unmodified_pair(p))
2473                return;
2474
2475        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2476            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2477                return; /* no tree diffs in patch format */
2478
2479        run_diff(p, o);
2480}
2481
2482static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2483                            struct diffstat_t *diffstat)
2484{
2485        if (diff_unmodified_pair(p))
2486                return;
2487
2488        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2489            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2490                return; /* no tree diffs in patch format */
2491
2492        run_diffstat(p, o, diffstat);
2493}
2494
2495static void diff_flush_checkdiff(struct diff_filepair *p,
2496                struct diff_options *o)
2497{
2498        if (diff_unmodified_pair(p))
2499                return;
2500
2501        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2502            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2503                return; /* no tree diffs in patch format */
2504
2505        run_checkdiff(p, o);
2506}
2507
2508int diff_queue_is_empty(void)
2509{
2510        struct diff_queue_struct *q = &diff_queued_diff;
2511        int i;
2512        for (i = 0; i < q->nr; i++)
2513                if (!diff_unmodified_pair(q->queue[i]))
2514                        return 0;
2515        return 1;
2516}
2517
2518#if DIFF_DEBUG
2519void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2520{
2521        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2522                x, one ? one : "",
2523                s->path,
2524                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2525                s->mode,
2526                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2527        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2528                x, one ? one : "",
2529                s->size, s->xfrm_flags);
2530}
2531
2532void diff_debug_filepair(const struct diff_filepair *p, int i)
2533{
2534        diff_debug_filespec(p->one, i, "one");
2535        diff_debug_filespec(p->two, i, "two");
2536        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2537                p->score, p->status ? p->status : '?',
2538                p->one->rename_used, p->broken_pair);
2539}
2540
2541void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2542{
2543        int i;
2544        if (msg)
2545                fprintf(stderr, "%s\n", msg);
2546        fprintf(stderr, "q->nr = %d\n", q->nr);
2547        for (i = 0; i < q->nr; i++) {
2548                struct diff_filepair *p = q->queue[i];
2549                diff_debug_filepair(p, i);
2550        }
2551}
2552#endif
2553
2554static void diff_resolve_rename_copy(void)
2555{
2556        int i;
2557        struct diff_filepair *p;
2558        struct diff_queue_struct *q = &diff_queued_diff;
2559
2560        diff_debug_queue("resolve-rename-copy", q);
2561
2562        for (i = 0; i < q->nr; i++) {
2563                p = q->queue[i];
2564                p->status = 0; /* undecided */
2565                if (DIFF_PAIR_UNMERGED(p))
2566                        p->status = DIFF_STATUS_UNMERGED;
2567                else if (!DIFF_FILE_VALID(p->one))
2568                        p->status = DIFF_STATUS_ADDED;
2569                else if (!DIFF_FILE_VALID(p->two))
2570                        p->status = DIFF_STATUS_DELETED;
2571                else if (DIFF_PAIR_TYPE_CHANGED(p))
2572                        p->status = DIFF_STATUS_TYPE_CHANGED;
2573
2574                /* from this point on, we are dealing with a pair
2575                 * whose both sides are valid and of the same type, i.e.
2576                 * either in-place edit or rename/copy edit.
2577                 */
2578                else if (DIFF_PAIR_RENAME(p)) {
2579                        /*
2580                         * A rename might have re-connected a broken
2581                         * pair up, causing the pathnames to be the
2582                         * same again. If so, that's not a rename at
2583                         * all, just a modification..
2584                         *
2585                         * Otherwise, see if this source was used for
2586                         * multiple renames, in which case we decrement
2587                         * the count, and call it a copy.
2588                         */
2589                        if (!strcmp(p->one->path, p->two->path))
2590                                p->status = DIFF_STATUS_MODIFIED;
2591                        else if (--p->one->rename_used > 0)
2592                                p->status = DIFF_STATUS_COPIED;
2593                        else
2594                                p->status = DIFF_STATUS_RENAMED;
2595                }
2596                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2597                         p->one->mode != p->two->mode ||
2598                         is_null_sha1(p->one->sha1))
2599                        p->status = DIFF_STATUS_MODIFIED;
2600                else {
2601                        /* This is a "no-change" entry and should not
2602                         * happen anymore, but prepare for broken callers.
2603                         */
2604                        error("feeding unmodified %s to diffcore",
2605                              p->one->path);
2606                        p->status = DIFF_STATUS_UNKNOWN;
2607                }
2608        }
2609        diff_debug_queue("resolve-rename-copy done", q);
2610}
2611
2612static int check_pair_status(struct diff_filepair *p)
2613{
2614        switch (p->status) {
2615        case DIFF_STATUS_UNKNOWN:
2616                return 0;
2617        case 0:
2618                die("internal error in diff-resolve-rename-copy");
2619        default:
2620                return 1;
2621        }
2622}
2623
2624static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2625{
2626        int fmt = opt->output_format;
2627
2628        if (fmt & DIFF_FORMAT_CHECKDIFF)
2629                diff_flush_checkdiff(p, opt);
2630        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2631                diff_flush_raw(p, opt);
2632        else if (fmt & DIFF_FORMAT_NAME)
2633                write_name_quoted(p->two->path, stdout, opt->line_termination);
2634}
2635
2636static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2637{
2638        if (fs->mode)
2639                printf(" %s mode %06o ", newdelete, fs->mode);
2640        else
2641                printf(" %s ", newdelete);
2642        write_name_quoted(fs->path, stdout, '\n');
2643}
2644
2645
2646static void show_mode_change(struct diff_filepair *p, int show_name)
2647{
2648        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2649                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2650                        show_name ? ' ' : '\n');
2651                if (show_name) {
2652                        write_name_quoted(p->two->path, stdout, '\n');
2653                }
2654        }
2655}
2656
2657static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2658{
2659        char *names = pprint_rename(p->one->path, p->two->path);
2660
2661        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2662        free(names);
2663        show_mode_change(p, 0);
2664}
2665
2666static void diff_summary(struct diff_filepair *p)
2667{
2668        switch(p->status) {
2669        case DIFF_STATUS_DELETED:
2670                show_file_mode_name("delete", p->one);
2671                break;
2672        case DIFF_STATUS_ADDED:
2673                show_file_mode_name("create", p->two);
2674                break;
2675        case DIFF_STATUS_COPIED:
2676                show_rename_copy("copy", p);
2677                break;
2678        case DIFF_STATUS_RENAMED:
2679                show_rename_copy("rename", p);
2680                break;
2681        default:
2682                if (p->score) {
2683                        fputs(" rewrite ", stdout);
2684                        write_name_quoted(p->two->path, stdout, ' ');
2685                        printf("(%d%%)\n", similarity_index(p));
2686                }
2687                show_mode_change(p, !p->score);
2688                break;
2689        }
2690}
2691
2692struct patch_id_t {
2693        struct xdiff_emit_state xm;
2694        SHA_CTX *ctx;
2695        int patchlen;
2696};
2697
2698static int remove_space(char *line, int len)
2699{
2700        int i;
2701        char *dst = line;
2702        unsigned char c;
2703
2704        for (i = 0; i < len; i++)
2705                if (!isspace((c = line[i])))
2706                        *dst++ = c;
2707
2708        return dst - line;
2709}
2710
2711static void patch_id_consume(void *priv, char *line, unsigned long len)
2712{
2713        struct patch_id_t *data = priv;
2714        int new_len;
2715
2716        /* Ignore line numbers when computing the SHA1 of the patch */
2717        if (!prefixcmp(line, "@@ -"))
2718                return;
2719
2720        new_len = remove_space(line, len);
2721
2722        SHA1_Update(data->ctx, line, new_len);
2723        data->patchlen += new_len;
2724}
2725
2726/* returns 0 upon success, and writes result into sha1 */
2727static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2728{
2729        struct diff_queue_struct *q = &diff_queued_diff;
2730        int i;
2731        SHA_CTX ctx;
2732        struct patch_id_t data;
2733        char buffer[PATH_MAX * 4 + 20];
2734
2735        SHA1_Init(&ctx);
2736        memset(&data, 0, sizeof(struct patch_id_t));
2737        data.ctx = &ctx;
2738        data.xm.consume = patch_id_consume;
2739
2740        for (i = 0; i < q->nr; i++) {
2741                xpparam_t xpp;
2742                xdemitconf_t xecfg;
2743                xdemitcb_t ecb;
2744                mmfile_t mf1, mf2;
2745                struct diff_filepair *p = q->queue[i];
2746                int len1, len2;
2747
2748                memset(&xecfg, 0, sizeof(xecfg));
2749                if (p->status == 0)
2750                        return error("internal diff status error");
2751                if (p->status == DIFF_STATUS_UNKNOWN)
2752                        continue;
2753                if (diff_unmodified_pair(p))
2754                        continue;
2755                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2756                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2757                        continue;
2758                if (DIFF_PAIR_UNMERGED(p))
2759                        continue;
2760
2761                diff_fill_sha1_info(p->one);
2762                diff_fill_sha1_info(p->two);
2763                if (fill_mmfile(&mf1, p->one) < 0 ||
2764                                fill_mmfile(&mf2, p->two) < 0)
2765                        return error("unable to read files to diff");
2766
2767                len1 = remove_space(p->one->path, strlen(p->one->path));
2768                len2 = remove_space(p->two->path, strlen(p->two->path));
2769                if (p->one->mode == 0)
2770                        len1 = snprintf(buffer, sizeof(buffer),
2771                                        "diff--gita/%.*sb/%.*s"
2772                                        "newfilemode%06o"
2773                                        "---/dev/null"
2774                                        "+++b/%.*s",
2775                                        len1, p->one->path,
2776                                        len2, p->two->path,
2777                                        p->two->mode,
2778                                        len2, p->two->path);
2779                else if (p->two->mode == 0)
2780                        len1 = snprintf(buffer, sizeof(buffer),
2781                                        "diff--gita/%.*sb/%.*s"
2782                                        "deletedfilemode%06o"
2783                                        "---a/%.*s"
2784                                        "+++/dev/null",
2785                                        len1, p->one->path,
2786                                        len2, p->two->path,
2787                                        p->one->mode,
2788                                        len1, p->one->path);
2789                else
2790                        len1 = snprintf(buffer, sizeof(buffer),
2791                                        "diff--gita/%.*sb/%.*s"
2792                                        "---a/%.*s"
2793                                        "+++b/%.*s",
2794                                        len1, p->one->path,
2795                                        len2, p->two->path,
2796                                        len1, p->one->path,
2797                                        len2, p->two->path);
2798                SHA1_Update(&ctx, buffer, len1);
2799
2800                xpp.flags = XDF_NEED_MINIMAL;
2801                xecfg.ctxlen = 3;
2802                xecfg.flags = XDL_EMIT_FUNCNAMES;
2803                ecb.outf = xdiff_outf;
2804                ecb.priv = &data;
2805                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2806        }
2807
2808        SHA1_Final(sha1, &ctx);
2809        return 0;
2810}
2811
2812int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2813{
2814        struct diff_queue_struct *q = &diff_queued_diff;
2815        int i;
2816        int result = diff_get_patch_id(options, sha1);
2817
2818        for (i = 0; i < q->nr; i++)
2819                diff_free_filepair(q->queue[i]);
2820
2821        free(q->queue);
2822        q->queue = NULL;
2823        q->nr = q->alloc = 0;
2824
2825        return result;
2826}
2827
2828static int is_summary_empty(const struct diff_queue_struct *q)
2829{
2830        int i;
2831
2832        for (i = 0; i < q->nr; i++) {
2833                const struct diff_filepair *p = q->queue[i];
2834
2835                switch (p->status) {
2836                case DIFF_STATUS_DELETED:
2837                case DIFF_STATUS_ADDED:
2838                case DIFF_STATUS_COPIED:
2839                case DIFF_STATUS_RENAMED:
2840                        return 0;
2841                default:
2842                        if (p->score)
2843                                return 0;
2844                        if (p->one->mode && p->two->mode &&
2845                            p->one->mode != p->two->mode)
2846                                return 0;
2847                        break;
2848                }
2849        }
2850        return 1;
2851}
2852
2853void diff_flush(struct diff_options *options)
2854{
2855        struct diff_queue_struct *q = &diff_queued_diff;
2856        int i, output_format = options->output_format;
2857        int separator = 0;
2858
2859        /*
2860         * Order: raw, stat, summary, patch
2861         * or:    name/name-status/checkdiff (other bits clear)
2862         */
2863        if (!q->nr)
2864                goto free_queue;
2865
2866        if (output_format & (DIFF_FORMAT_RAW |
2867                             DIFF_FORMAT_NAME |
2868                             DIFF_FORMAT_NAME_STATUS |
2869                             DIFF_FORMAT_CHECKDIFF)) {
2870                for (i = 0; i < q->nr; i++) {
2871                        struct diff_filepair *p = q->queue[i];
2872                        if (check_pair_status(p))
2873                                flush_one_pair(p, options);
2874                }
2875                separator++;
2876        }
2877
2878        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2879                struct diffstat_t diffstat;
2880
2881                memset(&diffstat, 0, sizeof(struct diffstat_t));
2882                diffstat.xm.consume = diffstat_consume;
2883                for (i = 0; i < q->nr; i++) {
2884                        struct diff_filepair *p = q->queue[i];
2885                        if (check_pair_status(p))
2886                                diff_flush_stat(p, options, &diffstat);
2887                }
2888                if (output_format & DIFF_FORMAT_NUMSTAT)
2889                        show_numstat(&diffstat, options);
2890                if (output_format & DIFF_FORMAT_DIFFSTAT)
2891                        show_stats(&diffstat, options);
2892                if (output_format & DIFF_FORMAT_SHORTSTAT)
2893                        show_shortstats(&diffstat);
2894                free_diffstat_info(&diffstat);
2895                separator++;
2896        }
2897
2898        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2899                for (i = 0; i < q->nr; i++)
2900                        diff_summary(q->queue[i]);
2901                separator++;
2902        }
2903
2904        if (output_format & DIFF_FORMAT_PATCH) {
2905                if (separator) {
2906                        if (options->stat_sep) {
2907                                /* attach patch instead of inline */
2908                                fputs(options->stat_sep, stdout);
2909                        } else {
2910                                putchar(options->line_termination);
2911                        }
2912                }
2913
2914                for (i = 0; i < q->nr; i++) {
2915                        struct diff_filepair *p = q->queue[i];
2916                        if (check_pair_status(p))
2917                                diff_flush_patch(p, options);
2918                }
2919        }
2920
2921        if (output_format & DIFF_FORMAT_CALLBACK)
2922                options->format_callback(q, options, options->format_callback_data);
2923
2924        for (i = 0; i < q->nr; i++)
2925                diff_free_filepair(q->queue[i]);
2926free_queue:
2927        free(q->queue);
2928        q->queue = NULL;
2929        q->nr = q->alloc = 0;
2930}
2931
2932static void diffcore_apply_filter(const char *filter)
2933{
2934        int i;
2935        struct diff_queue_struct *q = &diff_queued_diff;
2936        struct diff_queue_struct outq;
2937        outq.queue = NULL;
2938        outq.nr = outq.alloc = 0;
2939
2940        if (!filter)
2941                return;
2942
2943        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2944                int found;
2945                for (i = found = 0; !found && i < q->nr; i++) {
2946                        struct diff_filepair *p = q->queue[i];
2947                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2948                             ((p->score &&
2949                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2950                              (!p->score &&
2951                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2952                            ((p->status != DIFF_STATUS_MODIFIED) &&
2953                             strchr(filter, p->status)))
2954                                found++;
2955                }
2956                if (found)
2957                        return;
2958
2959                /* otherwise we will clear the whole queue
2960                 * by copying the empty outq at the end of this
2961                 * function, but first clear the current entries
2962                 * in the queue.
2963                 */
2964                for (i = 0; i < q->nr; i++)
2965                        diff_free_filepair(q->queue[i]);
2966        }
2967        else {
2968                /* Only the matching ones */
2969                for (i = 0; i < q->nr; i++) {
2970                        struct diff_filepair *p = q->queue[i];
2971
2972                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2973                             ((p->score &&
2974                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2975                              (!p->score &&
2976                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2977                            ((p->status != DIFF_STATUS_MODIFIED) &&
2978                             strchr(filter, p->status)))
2979                                diff_q(&outq, p);
2980                        else
2981                                diff_free_filepair(p);
2982                }
2983        }
2984        free(q->queue);
2985        *q = outq;
2986}
2987
2988/* Check whether two filespecs with the same mode and size are identical */
2989static int diff_filespec_is_identical(struct diff_filespec *one,
2990                                      struct diff_filespec *two)
2991{
2992        if (S_ISGITLINK(one->mode)) {
2993                diff_fill_sha1_info(one);
2994                diff_fill_sha1_info(two);
2995                return !hashcmp(one->sha1, two->sha1);
2996        }
2997        if (diff_populate_filespec(one, 0))
2998                return 0;
2999        if (diff_populate_filespec(two, 0))
3000                return 0;
3001        return !memcmp(one->data, two->data, one->size);
3002}
3003
3004static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3005{
3006        int i;
3007        struct diff_queue_struct *q = &diff_queued_diff;
3008        struct diff_queue_struct outq;
3009        outq.queue = NULL;
3010        outq.nr = outq.alloc = 0;
3011
3012        for (i = 0; i < q->nr; i++) {
3013                struct diff_filepair *p = q->queue[i];
3014
3015                /*
3016                 * 1. Entries that come from stat info dirtyness
3017                 *    always have both sides (iow, not create/delete),
3018                 *    one side of the object name is unknown, with
3019                 *    the same mode and size.  Keep the ones that
3020                 *    do not match these criteria.  They have real
3021                 *    differences.
3022                 *
3023                 * 2. At this point, the file is known to be modified,
3024                 *    with the same mode and size, and the object
3025                 *    name of one side is unknown.  Need to inspect
3026                 *    the identical contents.
3027                 */
3028                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3029                    !DIFF_FILE_VALID(p->two) ||
3030                    (p->one->sha1_valid && p->two->sha1_valid) ||
3031                    (p->one->mode != p->two->mode) ||
3032                    diff_populate_filespec(p->one, 1) ||
3033                    diff_populate_filespec(p->two, 1) ||
3034                    (p->one->size != p->two->size) ||
3035                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3036                        diff_q(&outq, p);
3037                else {
3038                        /*
3039                         * The caller can subtract 1 from skip_stat_unmatch
3040                         * to determine how many paths were dirty only
3041                         * due to stat info mismatch.
3042                         */
3043                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3044                                diffopt->skip_stat_unmatch++;
3045                        diff_free_filepair(p);
3046                }
3047        }
3048        free(q->queue);
3049        *q = outq;
3050}
3051
3052void diffcore_std(struct diff_options *options)
3053{
3054        if (DIFF_OPT_TST(options, QUIET))
3055                return;
3056
3057        if (options->skip_stat_unmatch && !DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3058                diffcore_skip_stat_unmatch(options);
3059        if (options->break_opt != -1)
3060                diffcore_break(options->break_opt);
3061        if (options->detect_rename)
3062                diffcore_rename(options);
3063        if (options->break_opt != -1)
3064                diffcore_merge_broken();
3065        if (options->pickaxe)
3066                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3067        if (options->orderfile)
3068                diffcore_order(options->orderfile);
3069        diff_resolve_rename_copy();
3070        diffcore_apply_filter(options->filter);
3071
3072        if (diff_queued_diff.nr)
3073                DIFF_OPT_SET(options, HAS_CHANGES);
3074        else
3075                DIFF_OPT_CLR(options, HAS_CHANGES);
3076}
3077
3078int diff_result_code(struct diff_options *opt, int status)
3079{
3080        int result = 0;
3081        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3082            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3083                return status;
3084        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3085            DIFF_OPT_TST(opt, HAS_CHANGES))
3086                result |= 01;
3087        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3088            DIFF_OPT_TST(opt, CHECK_FAILED))
3089                result |= 02;
3090        return result;
3091}
3092
3093void diff_addremove(struct diff_options *options,
3094                    int addremove, unsigned mode,
3095                    const unsigned char *sha1,
3096                    const char *base, const char *path)
3097{
3098        char concatpath[PATH_MAX];
3099        struct diff_filespec *one, *two;
3100
3101        /* This may look odd, but it is a preparation for
3102         * feeding "there are unchanged files which should
3103         * not produce diffs, but when you are doing copy
3104         * detection you would need them, so here they are"
3105         * entries to the diff-core.  They will be prefixed
3106         * with something like '=' or '*' (I haven't decided
3107         * which but should not make any difference).
3108         * Feeding the same new and old to diff_change()
3109         * also has the same effect.
3110         * Before the final output happens, they are pruned after
3111         * merged into rename/copy pairs as appropriate.
3112         */
3113        if (DIFF_OPT_TST(options, REVERSE_DIFF))
3114                addremove = (addremove == '+' ? '-' :
3115                             addremove == '-' ? '+' : addremove);
3116
3117        if (!path) path = "";
3118        sprintf(concatpath, "%s%s", base, path);
3119        one = alloc_filespec(concatpath);
3120        two = alloc_filespec(concatpath);
3121
3122        if (addremove != '+')
3123                fill_filespec(one, sha1, mode);
3124        if (addremove != '-')
3125                fill_filespec(two, sha1, mode);
3126
3127        diff_queue(&diff_queued_diff, one, two);
3128        DIFF_OPT_SET(options, HAS_CHANGES);
3129}
3130
3131void diff_change(struct diff_options *options,
3132                 unsigned old_mode, unsigned new_mode,
3133                 const unsigned char *old_sha1,
3134                 const unsigned char *new_sha1,
3135                 const char *base, const char *path)
3136{
3137        char concatpath[PATH_MAX];
3138        struct diff_filespec *one, *two;
3139
3140        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3141                unsigned tmp;
3142                const unsigned char *tmp_c;
3143                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3144                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3145        }
3146        if (!path) path = "";
3147        sprintf(concatpath, "%s%s", base, path);
3148        one = alloc_filespec(concatpath);
3149        two = alloc_filespec(concatpath);
3150        fill_filespec(one, old_sha1, old_mode);
3151        fill_filespec(two, new_sha1, new_mode);
3152
3153        diff_queue(&diff_queued_diff, one, two);
3154        DIFF_OPT_SET(options, HAS_CHANGES);
3155}
3156
3157void diff_unmerge(struct diff_options *options,
3158                  const char *path,
3159                  unsigned mode, const unsigned char *sha1)
3160{
3161        struct diff_filespec *one, *two;
3162        one = alloc_filespec(path);
3163        two = alloc_filespec(path);
3164        fill_filespec(one, sha1, mode);
3165        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3166}