diff.con commit Merge branch 'maint' of git://linux-nfs.org/~bfields/git (c477553)
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13
  14#ifdef NO_FAST_WORKING_DIRECTORY
  15#define FAST_WORKING_DIRECTORY 0
  16#else
  17#define FAST_WORKING_DIRECTORY 1
  18#endif
  19
  20static int diff_detect_rename_default;
  21static int diff_rename_limit_default = 100;
  22static int diff_use_color_default;
  23static const char *external_diff_cmd_cfg;
  24int diff_auto_refresh_index = 1;
  25
  26static char diff_colors[][COLOR_MAXLEN] = {
  27        "\033[m",       /* reset */
  28        "",             /* PLAIN (normal) */
  29        "\033[1m",      /* METAINFO (bold) */
  30        "\033[36m",     /* FRAGINFO (cyan) */
  31        "\033[31m",     /* OLD (red) */
  32        "\033[32m",     /* NEW (green) */
  33        "\033[33m",     /* COMMIT (yellow) */
  34        "\033[41m",     /* WHITESPACE (red background) */
  35};
  36
  37static int parse_diff_color_slot(const char *var, int ofs)
  38{
  39        if (!strcasecmp(var+ofs, "plain"))
  40                return DIFF_PLAIN;
  41        if (!strcasecmp(var+ofs, "meta"))
  42                return DIFF_METAINFO;
  43        if (!strcasecmp(var+ofs, "frag"))
  44                return DIFF_FRAGINFO;
  45        if (!strcasecmp(var+ofs, "old"))
  46                return DIFF_FILE_OLD;
  47        if (!strcasecmp(var+ofs, "new"))
  48                return DIFF_FILE_NEW;
  49        if (!strcasecmp(var+ofs, "commit"))
  50                return DIFF_COMMIT;
  51        if (!strcasecmp(var+ofs, "whitespace"))
  52                return DIFF_WHITESPACE;
  53        die("bad config variable '%s'", var);
  54}
  55
  56static struct ll_diff_driver {
  57        const char *name;
  58        struct ll_diff_driver *next;
  59        char *cmd;
  60} *user_diff, **user_diff_tail;
  61
  62static void read_config_if_needed(void)
  63{
  64        if (!user_diff_tail) {
  65                user_diff_tail = &user_diff;
  66                git_config(git_diff_ui_config);
  67        }
  68}
  69
  70/*
  71 * Currently there is only "diff.<drivername>.command" variable;
  72 * because there are "diff.color.<slot>" variables, we are parsing
  73 * this in a bit convoluted way to allow low level diff driver
  74 * called "color".
  75 */
  76static int parse_lldiff_command(const char *var, const char *ep, const char *value)
  77{
  78        const char *name;
  79        int namelen;
  80        struct ll_diff_driver *drv;
  81
  82        name = var + 5;
  83        namelen = ep - name;
  84        for (drv = user_diff; drv; drv = drv->next)
  85                if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
  86                        break;
  87        if (!drv) {
  88                drv = xcalloc(1, sizeof(struct ll_diff_driver));
  89                drv->name = xmemdupz(name, namelen);
  90                if (!user_diff_tail)
  91                        user_diff_tail = &user_diff;
  92                *user_diff_tail = drv;
  93                user_diff_tail = &(drv->next);
  94        }
  95
  96        if (!value)
  97                return error("%s: lacks value", var);
  98        drv->cmd = strdup(value);
  99        return 0;
 100}
 101
 102/*
 103 * 'diff.<what>.funcname' attribute can be specified in the configuration
 104 * to define a customized regexp to find the beginning of a function to
 105 * be used for hunk header lines of "diff -p" style output.
 106 */
 107static struct funcname_pattern {
 108        char *name;
 109        char *pattern;
 110        struct funcname_pattern *next;
 111} *funcname_pattern_list;
 112
 113static int parse_funcname_pattern(const char *var, const char *ep, const char *value)
 114{
 115        const char *name;
 116        int namelen;
 117        struct funcname_pattern *pp;
 118
 119        name = var + 5; /* "diff." */
 120        namelen = ep - name;
 121
 122        for (pp = funcname_pattern_list; pp; pp = pp->next)
 123                if (!strncmp(pp->name, name, namelen) && !pp->name[namelen])
 124                        break;
 125        if (!pp) {
 126                pp = xcalloc(1, sizeof(*pp));
 127                pp->name = xmemdupz(name, namelen);
 128                pp->next = funcname_pattern_list;
 129                funcname_pattern_list = pp;
 130        }
 131        if (pp->pattern)
 132                free(pp->pattern);
 133        pp->pattern = xstrdup(value);
 134        return 0;
 135}
 136
 137/*
 138 * These are to give UI layer defaults.
 139 * The core-level commands such as git-diff-files should
 140 * never be affected by the setting of diff.renames
 141 * the user happens to have in the configuration file.
 142 */
 143int git_diff_ui_config(const char *var, const char *value)
 144{
 145        if (!strcmp(var, "diff.renamelimit")) {
 146                diff_rename_limit_default = git_config_int(var, value);
 147                return 0;
 148        }
 149        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 150                diff_use_color_default = git_config_colorbool(var, value, -1);
 151                return 0;
 152        }
 153        if (!strcmp(var, "diff.renames")) {
 154                if (!value)
 155                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 156                else if (!strcasecmp(value, "copies") ||
 157                         !strcasecmp(value, "copy"))
 158                        diff_detect_rename_default = DIFF_DETECT_COPY;
 159                else if (git_config_bool(var,value))
 160                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 161                return 0;
 162        }
 163        if (!strcmp(var, "diff.autorefreshindex")) {
 164                diff_auto_refresh_index = git_config_bool(var, value);
 165                return 0;
 166        }
 167        if (!strcmp(var, "diff.external")) {
 168                external_diff_cmd_cfg = xstrdup(value);
 169                return 0;
 170        }
 171        if (!prefixcmp(var, "diff.")) {
 172                const char *ep = strrchr(var, '.');
 173
 174                if (ep != var + 4) {
 175                        if (!strcmp(ep, ".command"))
 176                                return parse_lldiff_command(var, ep, value);
 177                        if (!strcmp(ep, ".funcname"))
 178                                return parse_funcname_pattern(var, ep, value);
 179                }
 180        }
 181        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 182                int slot = parse_diff_color_slot(var, 11);
 183                color_parse(value, var, diff_colors[slot]);
 184                return 0;
 185        }
 186
 187        return git_default_config(var, value);
 188}
 189
 190static char *quote_two(const char *one, const char *two)
 191{
 192        int need_one = quote_c_style(one, NULL, NULL, 1);
 193        int need_two = quote_c_style(two, NULL, NULL, 1);
 194        struct strbuf res;
 195
 196        strbuf_init(&res, 0);
 197        if (need_one + need_two) {
 198                strbuf_addch(&res, '"');
 199                quote_c_style(one, &res, NULL, 1);
 200                quote_c_style(two, &res, NULL, 1);
 201                strbuf_addch(&res, '"');
 202        } else {
 203                strbuf_addstr(&res, one);
 204                strbuf_addstr(&res, two);
 205        }
 206        return strbuf_detach(&res, NULL);
 207}
 208
 209static const char *external_diff(void)
 210{
 211        static const char *external_diff_cmd = NULL;
 212        static int done_preparing = 0;
 213
 214        if (done_preparing)
 215                return external_diff_cmd;
 216        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 217        if (!external_diff_cmd)
 218                external_diff_cmd = external_diff_cmd_cfg;
 219        done_preparing = 1;
 220        return external_diff_cmd;
 221}
 222
 223static struct diff_tempfile {
 224        const char *name; /* filename external diff should read from */
 225        char hex[41];
 226        char mode[10];
 227        char tmp_path[PATH_MAX];
 228} diff_temp[2];
 229
 230static int count_lines(const char *data, int size)
 231{
 232        int count, ch, completely_empty = 1, nl_just_seen = 0;
 233        count = 0;
 234        while (0 < size--) {
 235                ch = *data++;
 236                if (ch == '\n') {
 237                        count++;
 238                        nl_just_seen = 1;
 239                        completely_empty = 0;
 240                }
 241                else {
 242                        nl_just_seen = 0;
 243                        completely_empty = 0;
 244                }
 245        }
 246        if (completely_empty)
 247                return 0;
 248        if (!nl_just_seen)
 249                count++; /* no trailing newline */
 250        return count;
 251}
 252
 253static void print_line_count(int count)
 254{
 255        switch (count) {
 256        case 0:
 257                printf("0,0");
 258                break;
 259        case 1:
 260                printf("1");
 261                break;
 262        default:
 263                printf("1,%d", count);
 264                break;
 265        }
 266}
 267
 268static void copy_file(int prefix, const char *data, int size,
 269                const char *set, const char *reset)
 270{
 271        int ch, nl_just_seen = 1;
 272        while (0 < size--) {
 273                ch = *data++;
 274                if (nl_just_seen) {
 275                        fputs(set, stdout);
 276                        putchar(prefix);
 277                }
 278                if (ch == '\n') {
 279                        nl_just_seen = 1;
 280                        fputs(reset, stdout);
 281                } else
 282                        nl_just_seen = 0;
 283                putchar(ch);
 284        }
 285        if (!nl_just_seen)
 286                printf("%s\n\\ No newline at end of file\n", reset);
 287}
 288
 289static void emit_rewrite_diff(const char *name_a,
 290                              const char *name_b,
 291                              struct diff_filespec *one,
 292                              struct diff_filespec *two,
 293                              struct diff_options *o)
 294{
 295        int lc_a, lc_b;
 296        int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
 297        const char *name_a_tab, *name_b_tab;
 298        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 299        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 300        const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
 301        const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
 302        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 303        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 304
 305        name_a += (*name_a == '/');
 306        name_b += (*name_b == '/');
 307        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 308        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 309
 310        strbuf_reset(&a_name);
 311        strbuf_reset(&b_name);
 312        quote_two_c_style(&a_name, o->a_prefix, name_a, 0);
 313        quote_two_c_style(&b_name, o->b_prefix, name_b, 0);
 314
 315        diff_populate_filespec(one, 0);
 316        diff_populate_filespec(two, 0);
 317        lc_a = count_lines(one->data, one->size);
 318        lc_b = count_lines(two->data, two->size);
 319        printf("%s--- %s%s%s\n%s+++ %s%s%s\n%s@@ -",
 320               metainfo, a_name.buf, name_a_tab, reset,
 321               metainfo, b_name.buf, name_b_tab, reset, fraginfo);
 322        print_line_count(lc_a);
 323        printf(" +");
 324        print_line_count(lc_b);
 325        printf(" @@%s\n", reset);
 326        if (lc_a)
 327                copy_file('-', one->data, one->size, old, reset);
 328        if (lc_b)
 329                copy_file('+', two->data, two->size, new, reset);
 330}
 331
 332static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 333{
 334        if (!DIFF_FILE_VALID(one)) {
 335                mf->ptr = (char *)""; /* does not matter */
 336                mf->size = 0;
 337                return 0;
 338        }
 339        else if (diff_populate_filespec(one, 0))
 340                return -1;
 341        mf->ptr = one->data;
 342        mf->size = one->size;
 343        return 0;
 344}
 345
 346struct diff_words_buffer {
 347        mmfile_t text;
 348        long alloc;
 349        long current; /* output pointer */
 350        int suppressed_newline;
 351};
 352
 353static void diff_words_append(char *line, unsigned long len,
 354                struct diff_words_buffer *buffer)
 355{
 356        if (buffer->text.size + len > buffer->alloc) {
 357                buffer->alloc = (buffer->text.size + len) * 3 / 2;
 358                buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
 359        }
 360        line++;
 361        len--;
 362        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 363        buffer->text.size += len;
 364}
 365
 366struct diff_words_data {
 367        struct xdiff_emit_state xm;
 368        struct diff_words_buffer minus, plus;
 369};
 370
 371static void print_word(struct diff_words_buffer *buffer, int len, int color,
 372                int suppress_newline)
 373{
 374        const char *ptr;
 375        int eol = 0;
 376
 377        if (len == 0)
 378                return;
 379
 380        ptr  = buffer->text.ptr + buffer->current;
 381        buffer->current += len;
 382
 383        if (ptr[len - 1] == '\n') {
 384                eol = 1;
 385                len--;
 386        }
 387
 388        fputs(diff_get_color(1, color), stdout);
 389        fwrite(ptr, len, 1, stdout);
 390        fputs(diff_get_color(1, DIFF_RESET), stdout);
 391
 392        if (eol) {
 393                if (suppress_newline)
 394                        buffer->suppressed_newline = 1;
 395                else
 396                        putchar('\n');
 397        }
 398}
 399
 400static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 401{
 402        struct diff_words_data *diff_words = priv;
 403
 404        if (diff_words->minus.suppressed_newline) {
 405                if (line[0] != '+')
 406                        putchar('\n');
 407                diff_words->minus.suppressed_newline = 0;
 408        }
 409
 410        len--;
 411        switch (line[0]) {
 412                case '-':
 413                        print_word(&diff_words->minus, len, DIFF_FILE_OLD, 1);
 414                        break;
 415                case '+':
 416                        print_word(&diff_words->plus, len, DIFF_FILE_NEW, 0);
 417                        break;
 418                case ' ':
 419                        print_word(&diff_words->plus, len, DIFF_PLAIN, 0);
 420                        diff_words->minus.current += len;
 421                        break;
 422        }
 423}
 424
 425/* this executes the word diff on the accumulated buffers */
 426static void diff_words_show(struct diff_words_data *diff_words)
 427{
 428        xpparam_t xpp;
 429        xdemitconf_t xecfg;
 430        xdemitcb_t ecb;
 431        mmfile_t minus, plus;
 432        int i;
 433
 434        memset(&xecfg, 0, sizeof(xecfg));
 435        minus.size = diff_words->minus.text.size;
 436        minus.ptr = xmalloc(minus.size);
 437        memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
 438        for (i = 0; i < minus.size; i++)
 439                if (isspace(minus.ptr[i]))
 440                        minus.ptr[i] = '\n';
 441        diff_words->minus.current = 0;
 442
 443        plus.size = diff_words->plus.text.size;
 444        plus.ptr = xmalloc(plus.size);
 445        memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
 446        for (i = 0; i < plus.size; i++)
 447                if (isspace(plus.ptr[i]))
 448                        plus.ptr[i] = '\n';
 449        diff_words->plus.current = 0;
 450
 451        xpp.flags = XDF_NEED_MINIMAL;
 452        xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
 453        ecb.outf = xdiff_outf;
 454        ecb.priv = diff_words;
 455        diff_words->xm.consume = fn_out_diff_words_aux;
 456        xdi_diff(&minus, &plus, &xpp, &xecfg, &ecb);
 457
 458        free(minus.ptr);
 459        free(plus.ptr);
 460        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 461
 462        if (diff_words->minus.suppressed_newline) {
 463                putchar('\n');
 464                diff_words->minus.suppressed_newline = 0;
 465        }
 466}
 467
 468struct emit_callback {
 469        struct xdiff_emit_state xm;
 470        int nparents, color_diff;
 471        unsigned ws_rule;
 472        const char **label_path;
 473        struct diff_words_data *diff_words;
 474        int *found_changesp;
 475};
 476
 477static void free_diff_words_data(struct emit_callback *ecbdata)
 478{
 479        if (ecbdata->diff_words) {
 480                /* flush buffers */
 481                if (ecbdata->diff_words->minus.text.size ||
 482                                ecbdata->diff_words->plus.text.size)
 483                        diff_words_show(ecbdata->diff_words);
 484
 485                if (ecbdata->diff_words->minus.text.ptr)
 486                        free (ecbdata->diff_words->minus.text.ptr);
 487                if (ecbdata->diff_words->plus.text.ptr)
 488                        free (ecbdata->diff_words->plus.text.ptr);
 489                free(ecbdata->diff_words);
 490                ecbdata->diff_words = NULL;
 491        }
 492}
 493
 494const char *diff_get_color(int diff_use_color, enum color_diff ix)
 495{
 496        if (diff_use_color)
 497                return diff_colors[ix];
 498        return "";
 499}
 500
 501static void emit_line(const char *set, const char *reset, const char *line, int len)
 502{
 503        fputs(set, stdout);
 504        fwrite(line, len, 1, stdout);
 505        fputs(reset, stdout);
 506}
 507
 508static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 509{
 510        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 511        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 512
 513        if (!*ws)
 514                emit_line(set, reset, line, len);
 515        else {
 516                /* Emit just the prefix, then the rest. */
 517                emit_line(set, reset, line, ecbdata->nparents);
 518                (void)check_and_emit_line(line + ecbdata->nparents,
 519                    len - ecbdata->nparents, ecbdata->ws_rule,
 520                    stdout, set, reset, ws);
 521        }
 522}
 523
 524static void fn_out_consume(void *priv, char *line, unsigned long len)
 525{
 526        int i;
 527        int color;
 528        struct emit_callback *ecbdata = priv;
 529        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 530        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 531
 532        *(ecbdata->found_changesp) = 1;
 533
 534        if (ecbdata->label_path[0]) {
 535                const char *name_a_tab, *name_b_tab;
 536
 537                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 538                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 539
 540                printf("%s--- %s%s%s\n",
 541                       set, ecbdata->label_path[0], reset, name_a_tab);
 542                printf("%s+++ %s%s%s\n",
 543                       set, ecbdata->label_path[1], reset, name_b_tab);
 544                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 545        }
 546
 547        /* This is not really necessary for now because
 548         * this codepath only deals with two-way diffs.
 549         */
 550        for (i = 0; i < len && line[i] == '@'; i++)
 551                ;
 552        if (2 <= i && i < len && line[i] == ' ') {
 553                ecbdata->nparents = i - 1;
 554                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 555                          reset, line, len);
 556                return;
 557        }
 558
 559        if (len < ecbdata->nparents) {
 560                set = reset;
 561                emit_line(reset, reset, line, len);
 562                return;
 563        }
 564
 565        color = DIFF_PLAIN;
 566        if (ecbdata->diff_words && ecbdata->nparents != 1)
 567                /* fall back to normal diff */
 568                free_diff_words_data(ecbdata);
 569        if (ecbdata->diff_words) {
 570                if (line[0] == '-') {
 571                        diff_words_append(line, len,
 572                                          &ecbdata->diff_words->minus);
 573                        return;
 574                } else if (line[0] == '+') {
 575                        diff_words_append(line, len,
 576                                          &ecbdata->diff_words->plus);
 577                        return;
 578                }
 579                if (ecbdata->diff_words->minus.text.size ||
 580                    ecbdata->diff_words->plus.text.size)
 581                        diff_words_show(ecbdata->diff_words);
 582                line++;
 583                len--;
 584                emit_line(set, reset, line, len);
 585                return;
 586        }
 587        for (i = 0; i < ecbdata->nparents && len; i++) {
 588                if (line[i] == '-')
 589                        color = DIFF_FILE_OLD;
 590                else if (line[i] == '+')
 591                        color = DIFF_FILE_NEW;
 592        }
 593
 594        if (color != DIFF_FILE_NEW) {
 595                emit_line(diff_get_color(ecbdata->color_diff, color),
 596                          reset, line, len);
 597                return;
 598        }
 599        emit_add_line(reset, ecbdata, line, len);
 600}
 601
 602static char *pprint_rename(const char *a, const char *b)
 603{
 604        const char *old = a;
 605        const char *new = b;
 606        struct strbuf name;
 607        int pfx_length, sfx_length;
 608        int len_a = strlen(a);
 609        int len_b = strlen(b);
 610        int a_midlen, b_midlen;
 611        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 612        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 613
 614        strbuf_init(&name, 0);
 615        if (qlen_a || qlen_b) {
 616                quote_c_style(a, &name, NULL, 0);
 617                strbuf_addstr(&name, " => ");
 618                quote_c_style(b, &name, NULL, 0);
 619                return strbuf_detach(&name, NULL);
 620        }
 621
 622        /* Find common prefix */
 623        pfx_length = 0;
 624        while (*old && *new && *old == *new) {
 625                if (*old == '/')
 626                        pfx_length = old - a + 1;
 627                old++;
 628                new++;
 629        }
 630
 631        /* Find common suffix */
 632        old = a + len_a;
 633        new = b + len_b;
 634        sfx_length = 0;
 635        while (a <= old && b <= new && *old == *new) {
 636                if (*old == '/')
 637                        sfx_length = len_a - (old - a);
 638                old--;
 639                new--;
 640        }
 641
 642        /*
 643         * pfx{mid-a => mid-b}sfx
 644         * {pfx-a => pfx-b}sfx
 645         * pfx{sfx-a => sfx-b}
 646         * name-a => name-b
 647         */
 648        a_midlen = len_a - pfx_length - sfx_length;
 649        b_midlen = len_b - pfx_length - sfx_length;
 650        if (a_midlen < 0)
 651                a_midlen = 0;
 652        if (b_midlen < 0)
 653                b_midlen = 0;
 654
 655        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 656        if (pfx_length + sfx_length) {
 657                strbuf_add(&name, a, pfx_length);
 658                strbuf_addch(&name, '{');
 659        }
 660        strbuf_add(&name, a + pfx_length, a_midlen);
 661        strbuf_addstr(&name, " => ");
 662        strbuf_add(&name, b + pfx_length, b_midlen);
 663        if (pfx_length + sfx_length) {
 664                strbuf_addch(&name, '}');
 665                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 666        }
 667        return strbuf_detach(&name, NULL);
 668}
 669
 670struct diffstat_t {
 671        struct xdiff_emit_state xm;
 672
 673        int nr;
 674        int alloc;
 675        struct diffstat_file {
 676                char *from_name;
 677                char *name;
 678                char *print_name;
 679                unsigned is_unmerged:1;
 680                unsigned is_binary:1;
 681                unsigned is_renamed:1;
 682                unsigned int added, deleted;
 683        } **files;
 684};
 685
 686static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 687                                          const char *name_a,
 688                                          const char *name_b)
 689{
 690        struct diffstat_file *x;
 691        x = xcalloc(sizeof (*x), 1);
 692        if (diffstat->nr == diffstat->alloc) {
 693                diffstat->alloc = alloc_nr(diffstat->alloc);
 694                diffstat->files = xrealloc(diffstat->files,
 695                                diffstat->alloc * sizeof(x));
 696        }
 697        diffstat->files[diffstat->nr++] = x;
 698        if (name_b) {
 699                x->from_name = xstrdup(name_a);
 700                x->name = xstrdup(name_b);
 701                x->is_renamed = 1;
 702        }
 703        else {
 704                x->from_name = NULL;
 705                x->name = xstrdup(name_a);
 706        }
 707        return x;
 708}
 709
 710static void diffstat_consume(void *priv, char *line, unsigned long len)
 711{
 712        struct diffstat_t *diffstat = priv;
 713        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 714
 715        if (line[0] == '+')
 716                x->added++;
 717        else if (line[0] == '-')
 718                x->deleted++;
 719}
 720
 721const char mime_boundary_leader[] = "------------";
 722
 723static int scale_linear(int it, int width, int max_change)
 724{
 725        /*
 726         * make sure that at least one '-' is printed if there were deletions,
 727         * and likewise for '+'.
 728         */
 729        if (max_change < 2)
 730                return it;
 731        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 732}
 733
 734static void show_name(const char *prefix, const char *name, int len,
 735                      const char *reset, const char *set)
 736{
 737        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 738}
 739
 740static void show_graph(char ch, int cnt, const char *set, const char *reset)
 741{
 742        if (cnt <= 0)
 743                return;
 744        printf("%s", set);
 745        while (cnt--)
 746                putchar(ch);
 747        printf("%s", reset);
 748}
 749
 750static void fill_print_name(struct diffstat_file *file)
 751{
 752        char *pname;
 753
 754        if (file->print_name)
 755                return;
 756
 757        if (!file->is_renamed) {
 758                struct strbuf buf;
 759                strbuf_init(&buf, 0);
 760                if (quote_c_style(file->name, &buf, NULL, 0)) {
 761                        pname = strbuf_detach(&buf, NULL);
 762                } else {
 763                        pname = file->name;
 764                        strbuf_release(&buf);
 765                }
 766        } else {
 767                pname = pprint_rename(file->from_name, file->name);
 768        }
 769        file->print_name = pname;
 770}
 771
 772static void show_stats(struct diffstat_t* data, struct diff_options *options)
 773{
 774        int i, len, add, del, total, adds = 0, dels = 0;
 775        int max_change = 0, max_len = 0;
 776        int total_files = data->nr;
 777        int width, name_width;
 778        const char *reset, *set, *add_c, *del_c;
 779
 780        if (data->nr == 0)
 781                return;
 782
 783        width = options->stat_width ? options->stat_width : 80;
 784        name_width = options->stat_name_width ? options->stat_name_width : 50;
 785
 786        /* Sanity: give at least 5 columns to the graph,
 787         * but leave at least 10 columns for the name.
 788         */
 789        if (width < name_width + 15) {
 790                if (name_width <= 25)
 791                        width = name_width + 15;
 792                else
 793                        name_width = width - 15;
 794        }
 795
 796        /* Find the longest filename and max number of changes */
 797        reset = diff_get_color_opt(options, DIFF_RESET);
 798        set   = diff_get_color_opt(options, DIFF_PLAIN);
 799        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
 800        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
 801
 802        for (i = 0; i < data->nr; i++) {
 803                struct diffstat_file *file = data->files[i];
 804                int change = file->added + file->deleted;
 805                fill_print_name(file);
 806                len = strlen(file->print_name);
 807                if (max_len < len)
 808                        max_len = len;
 809
 810                if (file->is_binary || file->is_unmerged)
 811                        continue;
 812                if (max_change < change)
 813                        max_change = change;
 814        }
 815
 816        /* Compute the width of the graph part;
 817         * 10 is for one blank at the beginning of the line plus
 818         * " | count " between the name and the graph.
 819         *
 820         * From here on, name_width is the width of the name area,
 821         * and width is the width of the graph area.
 822         */
 823        name_width = (name_width < max_len) ? name_width : max_len;
 824        if (width < (name_width + 10) + max_change)
 825                width = width - (name_width + 10);
 826        else
 827                width = max_change;
 828
 829        for (i = 0; i < data->nr; i++) {
 830                const char *prefix = "";
 831                char *name = data->files[i]->print_name;
 832                int added = data->files[i]->added;
 833                int deleted = data->files[i]->deleted;
 834                int name_len;
 835
 836                /*
 837                 * "scale" the filename
 838                 */
 839                len = name_width;
 840                name_len = strlen(name);
 841                if (name_width < name_len) {
 842                        char *slash;
 843                        prefix = "...";
 844                        len -= 3;
 845                        name += name_len - len;
 846                        slash = strchr(name, '/');
 847                        if (slash)
 848                                name = slash;
 849                }
 850
 851                if (data->files[i]->is_binary) {
 852                        show_name(prefix, name, len, reset, set);
 853                        printf("  Bin ");
 854                        printf("%s%d%s", del_c, deleted, reset);
 855                        printf(" -> ");
 856                        printf("%s%d%s", add_c, added, reset);
 857                        printf(" bytes");
 858                        printf("\n");
 859                        continue;
 860                }
 861                else if (data->files[i]->is_unmerged) {
 862                        show_name(prefix, name, len, reset, set);
 863                        printf("  Unmerged\n");
 864                        continue;
 865                }
 866                else if (!data->files[i]->is_renamed &&
 867                         (added + deleted == 0)) {
 868                        total_files--;
 869                        continue;
 870                }
 871
 872                /*
 873                 * scale the add/delete
 874                 */
 875                add = added;
 876                del = deleted;
 877                total = add + del;
 878                adds += add;
 879                dels += del;
 880
 881                if (width <= max_change) {
 882                        add = scale_linear(add, width, max_change);
 883                        del = scale_linear(del, width, max_change);
 884                        total = add + del;
 885                }
 886                show_name(prefix, name, len, reset, set);
 887                printf("%5d ", added + deleted);
 888                show_graph('+', add, add_c, reset);
 889                show_graph('-', del, del_c, reset);
 890                putchar('\n');
 891        }
 892        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 893               set, total_files, adds, dels, reset);
 894}
 895
 896static void show_shortstats(struct diffstat_t* data)
 897{
 898        int i, adds = 0, dels = 0, total_files = data->nr;
 899
 900        if (data->nr == 0)
 901                return;
 902
 903        for (i = 0; i < data->nr; i++) {
 904                if (!data->files[i]->is_binary &&
 905                    !data->files[i]->is_unmerged) {
 906                        int added = data->files[i]->added;
 907                        int deleted= data->files[i]->deleted;
 908                        if (!data->files[i]->is_renamed &&
 909                            (added + deleted == 0)) {
 910                                total_files--;
 911                        } else {
 912                                adds += added;
 913                                dels += deleted;
 914                        }
 915                }
 916        }
 917        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 918               total_files, adds, dels);
 919}
 920
 921static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 922{
 923        int i;
 924
 925        if (data->nr == 0)
 926                return;
 927
 928        for (i = 0; i < data->nr; i++) {
 929                struct diffstat_file *file = data->files[i];
 930
 931                if (file->is_binary)
 932                        printf("-\t-\t");
 933                else
 934                        printf("%d\t%d\t", file->added, file->deleted);
 935                if (options->line_termination) {
 936                        fill_print_name(file);
 937                        if (!file->is_renamed)
 938                                write_name_quoted(file->name, stdout,
 939                                                  options->line_termination);
 940                        else {
 941                                fputs(file->print_name, stdout);
 942                                putchar(options->line_termination);
 943                        }
 944                } else {
 945                        if (file->is_renamed) {
 946                                putchar('\0');
 947                                write_name_quoted(file->from_name, stdout, '\0');
 948                        }
 949                        write_name_quoted(file->name, stdout, '\0');
 950                }
 951        }
 952}
 953
 954static void free_diffstat_info(struct diffstat_t *diffstat)
 955{
 956        int i;
 957        for (i = 0; i < diffstat->nr; i++) {
 958                struct diffstat_file *f = diffstat->files[i];
 959                if (f->name != f->print_name)
 960                        free(f->print_name);
 961                free(f->name);
 962                free(f->from_name);
 963                free(f);
 964        }
 965        free(diffstat->files);
 966}
 967
 968struct checkdiff_t {
 969        struct xdiff_emit_state xm;
 970        const char *filename;
 971        int lineno, color_diff;
 972        unsigned ws_rule;
 973        unsigned status;
 974};
 975
 976static void checkdiff_consume(void *priv, char *line, unsigned long len)
 977{
 978        struct checkdiff_t *data = priv;
 979        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
 980        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
 981        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
 982        char *err;
 983
 984        if (line[0] == '+') {
 985                data->status = check_and_emit_line(line + 1, len - 1,
 986                    data->ws_rule, NULL, NULL, NULL, NULL);
 987                if (!data->status)
 988                        return;
 989                err = whitespace_error_string(data->status);
 990                printf("%s:%d: %s.\n", data->filename, data->lineno, err);
 991                free(err);
 992                emit_line(set, reset, line, 1);
 993                (void)check_and_emit_line(line + 1, len - 1, data->ws_rule,
 994                    stdout, set, reset, ws);
 995                data->lineno++;
 996        } else if (line[0] == ' ')
 997                data->lineno++;
 998        else if (line[0] == '@') {
 999                char *plus = strchr(line, '+');
1000                if (plus)
1001                        data->lineno = strtol(plus, NULL, 10);
1002                else
1003                        die("invalid diff");
1004        }
1005}
1006
1007static unsigned char *deflate_it(char *data,
1008                                 unsigned long size,
1009                                 unsigned long *result_size)
1010{
1011        int bound;
1012        unsigned char *deflated;
1013        z_stream stream;
1014
1015        memset(&stream, 0, sizeof(stream));
1016        deflateInit(&stream, zlib_compression_level);
1017        bound = deflateBound(&stream, size);
1018        deflated = xmalloc(bound);
1019        stream.next_out = deflated;
1020        stream.avail_out = bound;
1021
1022        stream.next_in = (unsigned char *)data;
1023        stream.avail_in = size;
1024        while (deflate(&stream, Z_FINISH) == Z_OK)
1025                ; /* nothing */
1026        deflateEnd(&stream);
1027        *result_size = stream.total_out;
1028        return deflated;
1029}
1030
1031static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1032{
1033        void *cp;
1034        void *delta;
1035        void *deflated;
1036        void *data;
1037        unsigned long orig_size;
1038        unsigned long delta_size;
1039        unsigned long deflate_size;
1040        unsigned long data_size;
1041
1042        /* We could do deflated delta, or we could do just deflated two,
1043         * whichever is smaller.
1044         */
1045        delta = NULL;
1046        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1047        if (one->size && two->size) {
1048                delta = diff_delta(one->ptr, one->size,
1049                                   two->ptr, two->size,
1050                                   &delta_size, deflate_size);
1051                if (delta) {
1052                        void *to_free = delta;
1053                        orig_size = delta_size;
1054                        delta = deflate_it(delta, delta_size, &delta_size);
1055                        free(to_free);
1056                }
1057        }
1058
1059        if (delta && delta_size < deflate_size) {
1060                printf("delta %lu\n", orig_size);
1061                free(deflated);
1062                data = delta;
1063                data_size = delta_size;
1064        }
1065        else {
1066                printf("literal %lu\n", two->size);
1067                free(delta);
1068                data = deflated;
1069                data_size = deflate_size;
1070        }
1071
1072        /* emit data encoded in base85 */
1073        cp = data;
1074        while (data_size) {
1075                int bytes = (52 < data_size) ? 52 : data_size;
1076                char line[70];
1077                data_size -= bytes;
1078                if (bytes <= 26)
1079                        line[0] = bytes + 'A' - 1;
1080                else
1081                        line[0] = bytes - 26 + 'a' - 1;
1082                encode_85(line + 1, cp, bytes);
1083                cp = (char *) cp + bytes;
1084                puts(line);
1085        }
1086        printf("\n");
1087        free(data);
1088}
1089
1090static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1091{
1092        printf("GIT binary patch\n");
1093        emit_binary_diff_body(one, two);
1094        emit_binary_diff_body(two, one);
1095}
1096
1097static void setup_diff_attr_check(struct git_attr_check *check)
1098{
1099        static struct git_attr *attr_diff;
1100
1101        if (!attr_diff) {
1102                attr_diff = git_attr("diff", 4);
1103        }
1104        check[0].attr = attr_diff;
1105}
1106
1107static void diff_filespec_check_attr(struct diff_filespec *one)
1108{
1109        struct git_attr_check attr_diff_check;
1110        int check_from_data = 0;
1111
1112        if (one->checked_attr)
1113                return;
1114
1115        setup_diff_attr_check(&attr_diff_check);
1116        one->is_binary = 0;
1117        one->funcname_pattern_ident = NULL;
1118
1119        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1120                const char *value;
1121
1122                /* binaryness */
1123                value = attr_diff_check.value;
1124                if (ATTR_TRUE(value))
1125                        ;
1126                else if (ATTR_FALSE(value))
1127                        one->is_binary = 1;
1128                else
1129                        check_from_data = 1;
1130
1131                /* funcname pattern ident */
1132                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1133                        ;
1134                else
1135                        one->funcname_pattern_ident = value;
1136        }
1137
1138        if (check_from_data) {
1139                if (!one->data && DIFF_FILE_VALID(one))
1140                        diff_populate_filespec(one, 0);
1141
1142                if (one->data)
1143                        one->is_binary = buffer_is_binary(one->data, one->size);
1144        }
1145}
1146
1147int diff_filespec_is_binary(struct diff_filespec *one)
1148{
1149        diff_filespec_check_attr(one);
1150        return one->is_binary;
1151}
1152
1153static const char *funcname_pattern(const char *ident)
1154{
1155        struct funcname_pattern *pp;
1156
1157        read_config_if_needed();
1158        for (pp = funcname_pattern_list; pp; pp = pp->next)
1159                if (!strcmp(ident, pp->name))
1160                        return pp->pattern;
1161        return NULL;
1162}
1163
1164static struct builtin_funcname_pattern {
1165        const char *name;
1166        const char *pattern;
1167} builtin_funcname_pattern[] = {
1168        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1169                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1170                        "^[     ]*\\(\\([       ]*"
1171                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1172                        "[      ]*([^;]*$\\)" },
1173        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1174};
1175
1176static const char *diff_funcname_pattern(struct diff_filespec *one)
1177{
1178        const char *ident, *pattern;
1179        int i;
1180
1181        diff_filespec_check_attr(one);
1182        ident = one->funcname_pattern_ident;
1183
1184        if (!ident)
1185                /*
1186                 * If the config file has "funcname.default" defined, that
1187                 * regexp is used; otherwise NULL is returned and xemit uses
1188                 * the built-in default.
1189                 */
1190                return funcname_pattern("default");
1191
1192        /* Look up custom "funcname.$ident" regexp from config. */
1193        pattern = funcname_pattern(ident);
1194        if (pattern)
1195                return pattern;
1196
1197        /*
1198         * And define built-in fallback patterns here.  Note that
1199         * these can be overriden by the user's config settings.
1200         */
1201        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1202                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1203                        return builtin_funcname_pattern[i].pattern;
1204
1205        return NULL;
1206}
1207
1208static void builtin_diff(const char *name_a,
1209                         const char *name_b,
1210                         struct diff_filespec *one,
1211                         struct diff_filespec *two,
1212                         const char *xfrm_msg,
1213                         struct diff_options *o,
1214                         int complete_rewrite)
1215{
1216        mmfile_t mf1, mf2;
1217        const char *lbl[2];
1218        char *a_one, *b_two;
1219        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1220        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1221
1222        a_one = quote_two(o->a_prefix, name_a + (*name_a == '/'));
1223        b_two = quote_two(o->b_prefix, name_b + (*name_b == '/'));
1224        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1225        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1226        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1227        if (lbl[0][0] == '/') {
1228                /* /dev/null */
1229                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1230                if (xfrm_msg && xfrm_msg[0])
1231                        printf("%s%s%s\n", set, xfrm_msg, reset);
1232        }
1233        else if (lbl[1][0] == '/') {
1234                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1235                if (xfrm_msg && xfrm_msg[0])
1236                        printf("%s%s%s\n", set, xfrm_msg, reset);
1237        }
1238        else {
1239                if (one->mode != two->mode) {
1240                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1241                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1242                }
1243                if (xfrm_msg && xfrm_msg[0])
1244                        printf("%s%s%s\n", set, xfrm_msg, reset);
1245                /*
1246                 * we do not run diff between different kind
1247                 * of objects.
1248                 */
1249                if ((one->mode ^ two->mode) & S_IFMT)
1250                        goto free_ab_and_return;
1251                if (complete_rewrite) {
1252                        emit_rewrite_diff(name_a, name_b, one, two, o);
1253                        o->found_changes = 1;
1254                        goto free_ab_and_return;
1255                }
1256        }
1257
1258        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1259                die("unable to read files to diff");
1260
1261        if (!DIFF_OPT_TST(o, TEXT) &&
1262            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1263                /* Quite common confusing case */
1264                if (mf1.size == mf2.size &&
1265                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1266                        goto free_ab_and_return;
1267                if (DIFF_OPT_TST(o, BINARY))
1268                        emit_binary_diff(&mf1, &mf2);
1269                else
1270                        printf("Binary files %s and %s differ\n",
1271                               lbl[0], lbl[1]);
1272                o->found_changes = 1;
1273        }
1274        else {
1275                /* Crazy xdl interfaces.. */
1276                const char *diffopts = getenv("GIT_DIFF_OPTS");
1277                xpparam_t xpp;
1278                xdemitconf_t xecfg;
1279                xdemitcb_t ecb;
1280                struct emit_callback ecbdata;
1281                const char *funcname_pattern;
1282
1283                funcname_pattern = diff_funcname_pattern(one);
1284                if (!funcname_pattern)
1285                        funcname_pattern = diff_funcname_pattern(two);
1286
1287                memset(&xecfg, 0, sizeof(xecfg));
1288                memset(&ecbdata, 0, sizeof(ecbdata));
1289                ecbdata.label_path = lbl;
1290                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1291                ecbdata.found_changesp = &o->found_changes;
1292                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1293                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1294                xecfg.ctxlen = o->context;
1295                xecfg.flags = XDL_EMIT_FUNCNAMES;
1296                if (funcname_pattern)
1297                        xdiff_set_find_func(&xecfg, funcname_pattern);
1298                if (!diffopts)
1299                        ;
1300                else if (!prefixcmp(diffopts, "--unified="))
1301                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1302                else if (!prefixcmp(diffopts, "-u"))
1303                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1304                ecb.outf = xdiff_outf;
1305                ecb.priv = &ecbdata;
1306                ecbdata.xm.consume = fn_out_consume;
1307                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1308                        ecbdata.diff_words =
1309                                xcalloc(1, sizeof(struct diff_words_data));
1310                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1311                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1312                        free_diff_words_data(&ecbdata);
1313        }
1314
1315 free_ab_and_return:
1316        diff_free_filespec_data(one);
1317        diff_free_filespec_data(two);
1318        free(a_one);
1319        free(b_two);
1320        return;
1321}
1322
1323static void builtin_diffstat(const char *name_a, const char *name_b,
1324                             struct diff_filespec *one,
1325                             struct diff_filespec *two,
1326                             struct diffstat_t *diffstat,
1327                             struct diff_options *o,
1328                             int complete_rewrite)
1329{
1330        mmfile_t mf1, mf2;
1331        struct diffstat_file *data;
1332
1333        data = diffstat_add(diffstat, name_a, name_b);
1334
1335        if (!one || !two) {
1336                data->is_unmerged = 1;
1337                return;
1338        }
1339        if (complete_rewrite) {
1340                diff_populate_filespec(one, 0);
1341                diff_populate_filespec(two, 0);
1342                data->deleted = count_lines(one->data, one->size);
1343                data->added = count_lines(two->data, two->size);
1344                goto free_and_return;
1345        }
1346        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1347                die("unable to read files to diff");
1348
1349        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1350                data->is_binary = 1;
1351                data->added = mf2.size;
1352                data->deleted = mf1.size;
1353        } else {
1354                /* Crazy xdl interfaces.. */
1355                xpparam_t xpp;
1356                xdemitconf_t xecfg;
1357                xdemitcb_t ecb;
1358
1359                memset(&xecfg, 0, sizeof(xecfg));
1360                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1361                ecb.outf = xdiff_outf;
1362                ecb.priv = diffstat;
1363                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1364        }
1365
1366 free_and_return:
1367        diff_free_filespec_data(one);
1368        diff_free_filespec_data(two);
1369}
1370
1371static void builtin_checkdiff(const char *name_a, const char *name_b,
1372                             struct diff_filespec *one,
1373                             struct diff_filespec *two, struct diff_options *o)
1374{
1375        mmfile_t mf1, mf2;
1376        struct checkdiff_t data;
1377
1378        if (!two)
1379                return;
1380
1381        memset(&data, 0, sizeof(data));
1382        data.xm.consume = checkdiff_consume;
1383        data.filename = name_b ? name_b : name_a;
1384        data.lineno = 0;
1385        data.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1386        data.ws_rule = whitespace_rule(data.filename);
1387
1388        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1389                die("unable to read files to diff");
1390
1391        if (diff_filespec_is_binary(two))
1392                goto free_and_return;
1393        else {
1394                /* Crazy xdl interfaces.. */
1395                xpparam_t xpp;
1396                xdemitconf_t xecfg;
1397                xdemitcb_t ecb;
1398
1399                memset(&xecfg, 0, sizeof(xecfg));
1400                xpp.flags = XDF_NEED_MINIMAL;
1401                ecb.outf = xdiff_outf;
1402                ecb.priv = &data;
1403                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1404        }
1405 free_and_return:
1406        diff_free_filespec_data(one);
1407        diff_free_filespec_data(two);
1408        if (data.status)
1409                DIFF_OPT_SET(o, CHECK_FAILED);
1410}
1411
1412struct diff_filespec *alloc_filespec(const char *path)
1413{
1414        int namelen = strlen(path);
1415        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1416
1417        memset(spec, 0, sizeof(*spec));
1418        spec->path = (char *)(spec + 1);
1419        memcpy(spec->path, path, namelen+1);
1420        spec->count = 1;
1421        return spec;
1422}
1423
1424void free_filespec(struct diff_filespec *spec)
1425{
1426        if (!--spec->count) {
1427                diff_free_filespec_data(spec);
1428                free(spec);
1429        }
1430}
1431
1432void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1433                   unsigned short mode)
1434{
1435        if (mode) {
1436                spec->mode = canon_mode(mode);
1437                hashcpy(spec->sha1, sha1);
1438                spec->sha1_valid = !is_null_sha1(sha1);
1439        }
1440}
1441
1442/*
1443 * Given a name and sha1 pair, if the index tells us the file in
1444 * the work tree has that object contents, return true, so that
1445 * prepare_temp_file() does not have to inflate and extract.
1446 */
1447static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1448{
1449        struct cache_entry *ce;
1450        struct stat st;
1451        int pos, len;
1452
1453        /* We do not read the cache ourselves here, because the
1454         * benchmark with my previous version that always reads cache
1455         * shows that it makes things worse for diff-tree comparing
1456         * two linux-2.6 kernel trees in an already checked out work
1457         * tree.  This is because most diff-tree comparisons deal with
1458         * only a small number of files, while reading the cache is
1459         * expensive for a large project, and its cost outweighs the
1460         * savings we get by not inflating the object to a temporary
1461         * file.  Practically, this code only helps when we are used
1462         * by diff-cache --cached, which does read the cache before
1463         * calling us.
1464         */
1465        if (!active_cache)
1466                return 0;
1467
1468        /* We want to avoid the working directory if our caller
1469         * doesn't need the data in a normal file, this system
1470         * is rather slow with its stat/open/mmap/close syscalls,
1471         * and the object is contained in a pack file.  The pack
1472         * is probably already open and will be faster to obtain
1473         * the data through than the working directory.  Loose
1474         * objects however would tend to be slower as they need
1475         * to be individually opened and inflated.
1476         */
1477        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1478                return 0;
1479
1480        len = strlen(name);
1481        pos = cache_name_pos(name, len);
1482        if (pos < 0)
1483                return 0;
1484        ce = active_cache[pos];
1485        if ((lstat(name, &st) < 0) ||
1486            !S_ISREG(st.st_mode) || /* careful! */
1487            ce_match_stat(ce, &st, 0) ||
1488            hashcmp(sha1, ce->sha1))
1489                return 0;
1490        /* we return 1 only when we can stat, it is a regular file,
1491         * stat information matches, and sha1 recorded in the cache
1492         * matches.  I.e. we know the file in the work tree really is
1493         * the same as the <name, sha1> pair.
1494         */
1495        return 1;
1496}
1497
1498static int populate_from_stdin(struct diff_filespec *s)
1499{
1500        struct strbuf buf;
1501        size_t size = 0;
1502
1503        strbuf_init(&buf, 0);
1504        if (strbuf_read(&buf, 0, 0) < 0)
1505                return error("error while reading from stdin %s",
1506                                     strerror(errno));
1507
1508        s->should_munmap = 0;
1509        s->data = strbuf_detach(&buf, &size);
1510        s->size = size;
1511        s->should_free = 1;
1512        return 0;
1513}
1514
1515static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1516{
1517        int len;
1518        char *data = xmalloc(100);
1519        len = snprintf(data, 100,
1520                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1521        s->data = data;
1522        s->size = len;
1523        s->should_free = 1;
1524        if (size_only) {
1525                s->data = NULL;
1526                free(data);
1527        }
1528        return 0;
1529}
1530
1531/*
1532 * While doing rename detection and pickaxe operation, we may need to
1533 * grab the data for the blob (or file) for our own in-core comparison.
1534 * diff_filespec has data and size fields for this purpose.
1535 */
1536int diff_populate_filespec(struct diff_filespec *s, int size_only)
1537{
1538        int err = 0;
1539        if (!DIFF_FILE_VALID(s))
1540                die("internal error: asking to populate invalid file.");
1541        if (S_ISDIR(s->mode))
1542                return -1;
1543
1544        if (s->data)
1545                return 0;
1546
1547        if (size_only && 0 < s->size)
1548                return 0;
1549
1550        if (S_ISGITLINK(s->mode))
1551                return diff_populate_gitlink(s, size_only);
1552
1553        if (!s->sha1_valid ||
1554            reuse_worktree_file(s->path, s->sha1, 0)) {
1555                struct strbuf buf;
1556                struct stat st;
1557                int fd;
1558
1559                if (!strcmp(s->path, "-"))
1560                        return populate_from_stdin(s);
1561
1562                if (lstat(s->path, &st) < 0) {
1563                        if (errno == ENOENT) {
1564                        err_empty:
1565                                err = -1;
1566                        empty:
1567                                s->data = (char *)"";
1568                                s->size = 0;
1569                                return err;
1570                        }
1571                }
1572                s->size = xsize_t(st.st_size);
1573                if (!s->size)
1574                        goto empty;
1575                if (size_only)
1576                        return 0;
1577                if (S_ISLNK(st.st_mode)) {
1578                        int ret;
1579                        s->data = xmalloc(s->size);
1580                        s->should_free = 1;
1581                        ret = readlink(s->path, s->data, s->size);
1582                        if (ret < 0) {
1583                                free(s->data);
1584                                goto err_empty;
1585                        }
1586                        return 0;
1587                }
1588                fd = open(s->path, O_RDONLY);
1589                if (fd < 0)
1590                        goto err_empty;
1591                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1592                close(fd);
1593                s->should_munmap = 1;
1594
1595                /*
1596                 * Convert from working tree format to canonical git format
1597                 */
1598                strbuf_init(&buf, 0);
1599                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1600                        size_t size = 0;
1601                        munmap(s->data, s->size);
1602                        s->should_munmap = 0;
1603                        s->data = strbuf_detach(&buf, &size);
1604                        s->size = size;
1605                        s->should_free = 1;
1606                }
1607        }
1608        else {
1609                enum object_type type;
1610                if (size_only)
1611                        type = sha1_object_info(s->sha1, &s->size);
1612                else {
1613                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1614                        s->should_free = 1;
1615                }
1616        }
1617        return 0;
1618}
1619
1620void diff_free_filespec_blob(struct diff_filespec *s)
1621{
1622        if (s->should_free)
1623                free(s->data);
1624        else if (s->should_munmap)
1625                munmap(s->data, s->size);
1626
1627        if (s->should_free || s->should_munmap) {
1628                s->should_free = s->should_munmap = 0;
1629                s->data = NULL;
1630        }
1631}
1632
1633void diff_free_filespec_data(struct diff_filespec *s)
1634{
1635        diff_free_filespec_blob(s);
1636        free(s->cnt_data);
1637        s->cnt_data = NULL;
1638}
1639
1640static void prep_temp_blob(struct diff_tempfile *temp,
1641                           void *blob,
1642                           unsigned long size,
1643                           const unsigned char *sha1,
1644                           int mode)
1645{
1646        int fd;
1647
1648        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1649        if (fd < 0)
1650                die("unable to create temp-file: %s", strerror(errno));
1651        if (write_in_full(fd, blob, size) != size)
1652                die("unable to write temp-file");
1653        close(fd);
1654        temp->name = temp->tmp_path;
1655        strcpy(temp->hex, sha1_to_hex(sha1));
1656        temp->hex[40] = 0;
1657        sprintf(temp->mode, "%06o", mode);
1658}
1659
1660static void prepare_temp_file(const char *name,
1661                              struct diff_tempfile *temp,
1662                              struct diff_filespec *one)
1663{
1664        if (!DIFF_FILE_VALID(one)) {
1665        not_a_valid_file:
1666                /* A '-' entry produces this for file-2, and
1667                 * a '+' entry produces this for file-1.
1668                 */
1669                temp->name = "/dev/null";
1670                strcpy(temp->hex, ".");
1671                strcpy(temp->mode, ".");
1672                return;
1673        }
1674
1675        if (!one->sha1_valid ||
1676            reuse_worktree_file(name, one->sha1, 1)) {
1677                struct stat st;
1678                if (lstat(name, &st) < 0) {
1679                        if (errno == ENOENT)
1680                                goto not_a_valid_file;
1681                        die("stat(%s): %s", name, strerror(errno));
1682                }
1683                if (S_ISLNK(st.st_mode)) {
1684                        int ret;
1685                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1686                        size_t sz = xsize_t(st.st_size);
1687                        if (sizeof(buf) <= st.st_size)
1688                                die("symlink too long: %s", name);
1689                        ret = readlink(name, buf, sz);
1690                        if (ret < 0)
1691                                die("readlink(%s)", name);
1692                        prep_temp_blob(temp, buf, sz,
1693                                       (one->sha1_valid ?
1694                                        one->sha1 : null_sha1),
1695                                       (one->sha1_valid ?
1696                                        one->mode : S_IFLNK));
1697                }
1698                else {
1699                        /* we can borrow from the file in the work tree */
1700                        temp->name = name;
1701                        if (!one->sha1_valid)
1702                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1703                        else
1704                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1705                        /* Even though we may sometimes borrow the
1706                         * contents from the work tree, we always want
1707                         * one->mode.  mode is trustworthy even when
1708                         * !(one->sha1_valid), as long as
1709                         * DIFF_FILE_VALID(one).
1710                         */
1711                        sprintf(temp->mode, "%06o", one->mode);
1712                }
1713                return;
1714        }
1715        else {
1716                if (diff_populate_filespec(one, 0))
1717                        die("cannot read data blob for %s", one->path);
1718                prep_temp_blob(temp, one->data, one->size,
1719                               one->sha1, one->mode);
1720        }
1721}
1722
1723static void remove_tempfile(void)
1724{
1725        int i;
1726
1727        for (i = 0; i < 2; i++)
1728                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1729                        unlink(diff_temp[i].name);
1730                        diff_temp[i].name = NULL;
1731                }
1732}
1733
1734static void remove_tempfile_on_signal(int signo)
1735{
1736        remove_tempfile();
1737        signal(SIGINT, SIG_DFL);
1738        raise(signo);
1739}
1740
1741/* An external diff command takes:
1742 *
1743 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1744 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1745 *
1746 */
1747static void run_external_diff(const char *pgm,
1748                              const char *name,
1749                              const char *other,
1750                              struct diff_filespec *one,
1751                              struct diff_filespec *two,
1752                              const char *xfrm_msg,
1753                              int complete_rewrite)
1754{
1755        const char *spawn_arg[10];
1756        struct diff_tempfile *temp = diff_temp;
1757        int retval;
1758        static int atexit_asked = 0;
1759        const char *othername;
1760        const char **arg = &spawn_arg[0];
1761
1762        othername = (other? other : name);
1763        if (one && two) {
1764                prepare_temp_file(name, &temp[0], one);
1765                prepare_temp_file(othername, &temp[1], two);
1766                if (! atexit_asked &&
1767                    (temp[0].name == temp[0].tmp_path ||
1768                     temp[1].name == temp[1].tmp_path)) {
1769                        atexit_asked = 1;
1770                        atexit(remove_tempfile);
1771                }
1772                signal(SIGINT, remove_tempfile_on_signal);
1773        }
1774
1775        if (one && two) {
1776                *arg++ = pgm;
1777                *arg++ = name;
1778                *arg++ = temp[0].name;
1779                *arg++ = temp[0].hex;
1780                *arg++ = temp[0].mode;
1781                *arg++ = temp[1].name;
1782                *arg++ = temp[1].hex;
1783                *arg++ = temp[1].mode;
1784                if (other) {
1785                        *arg++ = other;
1786                        *arg++ = xfrm_msg;
1787                }
1788        } else {
1789                *arg++ = pgm;
1790                *arg++ = name;
1791        }
1792        *arg = NULL;
1793        fflush(NULL);
1794        retval = run_command_v_opt(spawn_arg, 0);
1795        remove_tempfile();
1796        if (retval) {
1797                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1798                exit(1);
1799        }
1800}
1801
1802static const char *external_diff_attr(const char *name)
1803{
1804        struct git_attr_check attr_diff_check;
1805
1806        setup_diff_attr_check(&attr_diff_check);
1807        if (!git_checkattr(name, 1, &attr_diff_check)) {
1808                const char *value = attr_diff_check.value;
1809                if (!ATTR_TRUE(value) &&
1810                    !ATTR_FALSE(value) &&
1811                    !ATTR_UNSET(value)) {
1812                        struct ll_diff_driver *drv;
1813
1814                        read_config_if_needed();
1815                        for (drv = user_diff; drv; drv = drv->next)
1816                                if (!strcmp(drv->name, value))
1817                                        return drv->cmd;
1818                }
1819        }
1820        return NULL;
1821}
1822
1823static void run_diff_cmd(const char *pgm,
1824                         const char *name,
1825                         const char *other,
1826                         struct diff_filespec *one,
1827                         struct diff_filespec *two,
1828                         const char *xfrm_msg,
1829                         struct diff_options *o,
1830                         int complete_rewrite)
1831{
1832        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
1833                pgm = NULL;
1834        else {
1835                const char *cmd = external_diff_attr(name);
1836                if (cmd)
1837                        pgm = cmd;
1838        }
1839
1840        if (pgm) {
1841                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1842                                  complete_rewrite);
1843                return;
1844        }
1845        if (one && two)
1846                builtin_diff(name, other ? other : name,
1847                             one, two, xfrm_msg, o, complete_rewrite);
1848        else
1849                printf("* Unmerged path %s\n", name);
1850}
1851
1852static void diff_fill_sha1_info(struct diff_filespec *one)
1853{
1854        if (DIFF_FILE_VALID(one)) {
1855                if (!one->sha1_valid) {
1856                        struct stat st;
1857                        if (!strcmp(one->path, "-")) {
1858                                hashcpy(one->sha1, null_sha1);
1859                                return;
1860                        }
1861                        if (lstat(one->path, &st) < 0)
1862                                die("stat %s", one->path);
1863                        if (index_path(one->sha1, one->path, &st, 0))
1864                                die("cannot hash %s\n", one->path);
1865                }
1866        }
1867        else
1868                hashclr(one->sha1);
1869}
1870
1871static int similarity_index(struct diff_filepair *p)
1872{
1873        return p->score * 100 / MAX_SCORE;
1874}
1875
1876static void run_diff(struct diff_filepair *p, struct diff_options *o)
1877{
1878        const char *pgm = external_diff();
1879        struct strbuf msg;
1880        char *xfrm_msg;
1881        struct diff_filespec *one = p->one;
1882        struct diff_filespec *two = p->two;
1883        const char *name;
1884        const char *other;
1885        int complete_rewrite = 0;
1886
1887
1888        if (DIFF_PAIR_UNMERGED(p)) {
1889                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1890                return;
1891        }
1892
1893        name  = p->one->path;
1894        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1895        diff_fill_sha1_info(one);
1896        diff_fill_sha1_info(two);
1897
1898        strbuf_init(&msg, PATH_MAX * 2 + 300);
1899        switch (p->status) {
1900        case DIFF_STATUS_COPIED:
1901                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1902                strbuf_addstr(&msg, "\ncopy from ");
1903                quote_c_style(name, &msg, NULL, 0);
1904                strbuf_addstr(&msg, "\ncopy to ");
1905                quote_c_style(other, &msg, NULL, 0);
1906                strbuf_addch(&msg, '\n');
1907                break;
1908        case DIFF_STATUS_RENAMED:
1909                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1910                strbuf_addstr(&msg, "\nrename from ");
1911                quote_c_style(name, &msg, NULL, 0);
1912                strbuf_addstr(&msg, "\nrename to ");
1913                quote_c_style(other, &msg, NULL, 0);
1914                strbuf_addch(&msg, '\n');
1915                break;
1916        case DIFF_STATUS_MODIFIED:
1917                if (p->score) {
1918                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
1919                                        similarity_index(p));
1920                        complete_rewrite = 1;
1921                        break;
1922                }
1923                /* fallthru */
1924        default:
1925                /* nothing */
1926                ;
1927        }
1928
1929        if (hashcmp(one->sha1, two->sha1)) {
1930                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
1931
1932                if (DIFF_OPT_TST(o, BINARY)) {
1933                        mmfile_t mf;
1934                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
1935                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
1936                                abbrev = 40;
1937                }
1938                strbuf_addf(&msg, "index %.*s..%.*s",
1939                                abbrev, sha1_to_hex(one->sha1),
1940                                abbrev, sha1_to_hex(two->sha1));
1941                if (one->mode == two->mode)
1942                        strbuf_addf(&msg, " %06o", one->mode);
1943                strbuf_addch(&msg, '\n');
1944        }
1945
1946        if (msg.len)
1947                strbuf_setlen(&msg, msg.len - 1);
1948        xfrm_msg = msg.len ? msg.buf : NULL;
1949
1950        if (!pgm &&
1951            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1952            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1953                /* a filepair that changes between file and symlink
1954                 * needs to be split into deletion and creation.
1955                 */
1956                struct diff_filespec *null = alloc_filespec(two->path);
1957                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1958                free(null);
1959                null = alloc_filespec(one->path);
1960                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1961                free(null);
1962        }
1963        else
1964                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
1965                             complete_rewrite);
1966
1967        strbuf_release(&msg);
1968}
1969
1970static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
1971                         struct diffstat_t *diffstat)
1972{
1973        const char *name;
1974        const char *other;
1975        int complete_rewrite = 0;
1976
1977        if (DIFF_PAIR_UNMERGED(p)) {
1978                /* unmerged */
1979                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
1980                return;
1981        }
1982
1983        name = p->one->path;
1984        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1985
1986        diff_fill_sha1_info(p->one);
1987        diff_fill_sha1_info(p->two);
1988
1989        if (p->status == DIFF_STATUS_MODIFIED && p->score)
1990                complete_rewrite = 1;
1991        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
1992}
1993
1994static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
1995{
1996        const char *name;
1997        const char *other;
1998
1999        if (DIFF_PAIR_UNMERGED(p)) {
2000                /* unmerged */
2001                return;
2002        }
2003
2004        name = p->one->path;
2005        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2006
2007        diff_fill_sha1_info(p->one);
2008        diff_fill_sha1_info(p->two);
2009
2010        builtin_checkdiff(name, other, p->one, p->two, o);
2011}
2012
2013void diff_setup(struct diff_options *options)
2014{
2015        memset(options, 0, sizeof(*options));
2016        options->line_termination = '\n';
2017        options->break_opt = -1;
2018        options->rename_limit = -1;
2019        options->context = 3;
2020        options->msg_sep = "";
2021
2022        options->change = diff_change;
2023        options->add_remove = diff_addremove;
2024        if (diff_use_color_default)
2025                DIFF_OPT_SET(options, COLOR_DIFF);
2026        else
2027                DIFF_OPT_CLR(options, COLOR_DIFF);
2028        options->detect_rename = diff_detect_rename_default;
2029
2030        options->a_prefix = "a/";
2031        options->b_prefix = "b/";
2032}
2033
2034int diff_setup_done(struct diff_options *options)
2035{
2036        int count = 0;
2037
2038        if (options->output_format & DIFF_FORMAT_NAME)
2039                count++;
2040        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2041                count++;
2042        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2043                count++;
2044        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2045                count++;
2046        if (count > 1)
2047                die("--name-only, --name-status, --check and -s are mutually exclusive");
2048
2049        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2050                options->detect_rename = DIFF_DETECT_COPY;
2051
2052        if (options->output_format & (DIFF_FORMAT_NAME |
2053                                      DIFF_FORMAT_NAME_STATUS |
2054                                      DIFF_FORMAT_CHECKDIFF |
2055                                      DIFF_FORMAT_NO_OUTPUT))
2056                options->output_format &= ~(DIFF_FORMAT_RAW |
2057                                            DIFF_FORMAT_NUMSTAT |
2058                                            DIFF_FORMAT_DIFFSTAT |
2059                                            DIFF_FORMAT_SHORTSTAT |
2060                                            DIFF_FORMAT_SUMMARY |
2061                                            DIFF_FORMAT_PATCH);
2062
2063        /*
2064         * These cases always need recursive; we do not drop caller-supplied
2065         * recursive bits for other formats here.
2066         */
2067        if (options->output_format & (DIFF_FORMAT_PATCH |
2068                                      DIFF_FORMAT_NUMSTAT |
2069                                      DIFF_FORMAT_DIFFSTAT |
2070                                      DIFF_FORMAT_SHORTSTAT |
2071                                      DIFF_FORMAT_SUMMARY |
2072                                      DIFF_FORMAT_CHECKDIFF))
2073                DIFF_OPT_SET(options, RECURSIVE);
2074        /*
2075         * Also pickaxe would not work very well if you do not say recursive
2076         */
2077        if (options->pickaxe)
2078                DIFF_OPT_SET(options, RECURSIVE);
2079
2080        if (options->detect_rename && options->rename_limit < 0)
2081                options->rename_limit = diff_rename_limit_default;
2082        if (options->setup & DIFF_SETUP_USE_CACHE) {
2083                if (!active_cache)
2084                        /* read-cache does not die even when it fails
2085                         * so it is safe for us to do this here.  Also
2086                         * it does not smudge active_cache or active_nr
2087                         * when it fails, so we do not have to worry about
2088                         * cleaning it up ourselves either.
2089                         */
2090                        read_cache();
2091        }
2092        if (options->abbrev <= 0 || 40 < options->abbrev)
2093                options->abbrev = 40; /* full */
2094
2095        /*
2096         * It does not make sense to show the first hit we happened
2097         * to have found.  It does not make sense not to return with
2098         * exit code in such a case either.
2099         */
2100        if (DIFF_OPT_TST(options, QUIET)) {
2101                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2102                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2103        }
2104
2105        /*
2106         * If we postprocess in diffcore, we cannot simply return
2107         * upon the first hit.  We need to run diff as usual.
2108         */
2109        if (options->pickaxe || options->filter)
2110                DIFF_OPT_CLR(options, QUIET);
2111
2112        return 0;
2113}
2114
2115static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2116{
2117        char c, *eq;
2118        int len;
2119
2120        if (*arg != '-')
2121                return 0;
2122        c = *++arg;
2123        if (!c)
2124                return 0;
2125        if (c == arg_short) {
2126                c = *++arg;
2127                if (!c)
2128                        return 1;
2129                if (val && isdigit(c)) {
2130                        char *end;
2131                        int n = strtoul(arg, &end, 10);
2132                        if (*end)
2133                                return 0;
2134                        *val = n;
2135                        return 1;
2136                }
2137                return 0;
2138        }
2139        if (c != '-')
2140                return 0;
2141        arg++;
2142        eq = strchr(arg, '=');
2143        if (eq)
2144                len = eq - arg;
2145        else
2146                len = strlen(arg);
2147        if (!len || strncmp(arg, arg_long, len))
2148                return 0;
2149        if (eq) {
2150                int n;
2151                char *end;
2152                if (!isdigit(*++eq))
2153                        return 0;
2154                n = strtoul(eq, &end, 10);
2155                if (*end)
2156                        return 0;
2157                *val = n;
2158        }
2159        return 1;
2160}
2161
2162static int diff_scoreopt_parse(const char *opt);
2163
2164int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2165{
2166        const char *arg = av[0];
2167
2168        /* Output format options */
2169        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2170                options->output_format |= DIFF_FORMAT_PATCH;
2171        else if (opt_arg(arg, 'U', "unified", &options->context))
2172                options->output_format |= DIFF_FORMAT_PATCH;
2173        else if (!strcmp(arg, "--raw"))
2174                options->output_format |= DIFF_FORMAT_RAW;
2175        else if (!strcmp(arg, "--patch-with-raw"))
2176                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2177        else if (!strcmp(arg, "--numstat"))
2178                options->output_format |= DIFF_FORMAT_NUMSTAT;
2179        else if (!strcmp(arg, "--shortstat"))
2180                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2181        else if (!strcmp(arg, "--check"))
2182                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2183        else if (!strcmp(arg, "--summary"))
2184                options->output_format |= DIFF_FORMAT_SUMMARY;
2185        else if (!strcmp(arg, "--patch-with-stat"))
2186                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2187        else if (!strcmp(arg, "--name-only"))
2188                options->output_format |= DIFF_FORMAT_NAME;
2189        else if (!strcmp(arg, "--name-status"))
2190                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2191        else if (!strcmp(arg, "-s"))
2192                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2193        else if (!prefixcmp(arg, "--stat")) {
2194                char *end;
2195                int width = options->stat_width;
2196                int name_width = options->stat_name_width;
2197                arg += 6;
2198                end = (char *)arg;
2199
2200                switch (*arg) {
2201                case '-':
2202                        if (!prefixcmp(arg, "-width="))
2203                                width = strtoul(arg + 7, &end, 10);
2204                        else if (!prefixcmp(arg, "-name-width="))
2205                                name_width = strtoul(arg + 12, &end, 10);
2206                        break;
2207                case '=':
2208                        width = strtoul(arg+1, &end, 10);
2209                        if (*end == ',')
2210                                name_width = strtoul(end+1, &end, 10);
2211                }
2212
2213                /* Important! This checks all the error cases! */
2214                if (*end)
2215                        return 0;
2216                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2217                options->stat_name_width = name_width;
2218                options->stat_width = width;
2219        }
2220
2221        /* renames options */
2222        else if (!prefixcmp(arg, "-B")) {
2223                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2224                        return -1;
2225        }
2226        else if (!prefixcmp(arg, "-M")) {
2227                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2228                        return -1;
2229                options->detect_rename = DIFF_DETECT_RENAME;
2230        }
2231        else if (!prefixcmp(arg, "-C")) {
2232                if (options->detect_rename == DIFF_DETECT_COPY)
2233                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2234                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2235                        return -1;
2236                options->detect_rename = DIFF_DETECT_COPY;
2237        }
2238        else if (!strcmp(arg, "--no-renames"))
2239                options->detect_rename = 0;
2240
2241        /* xdiff options */
2242        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2243                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2244        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2245                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2246        else if (!strcmp(arg, "--ignore-space-at-eol"))
2247                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2248
2249        /* flags options */
2250        else if (!strcmp(arg, "--binary")) {
2251                options->output_format |= DIFF_FORMAT_PATCH;
2252                DIFF_OPT_SET(options, BINARY);
2253        }
2254        else if (!strcmp(arg, "--full-index"))
2255                DIFF_OPT_SET(options, FULL_INDEX);
2256        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2257                DIFF_OPT_SET(options, TEXT);
2258        else if (!strcmp(arg, "-R"))
2259                DIFF_OPT_SET(options, REVERSE_DIFF);
2260        else if (!strcmp(arg, "--find-copies-harder"))
2261                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2262        else if (!strcmp(arg, "--follow"))
2263                DIFF_OPT_SET(options, FOLLOW_RENAMES);
2264        else if (!strcmp(arg, "--color"))
2265                DIFF_OPT_SET(options, COLOR_DIFF);
2266        else if (!strcmp(arg, "--no-color"))
2267                DIFF_OPT_CLR(options, COLOR_DIFF);
2268        else if (!strcmp(arg, "--color-words"))
2269                options->flags |= DIFF_OPT_COLOR_DIFF | DIFF_OPT_COLOR_DIFF_WORDS;
2270        else if (!strcmp(arg, "--exit-code"))
2271                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2272        else if (!strcmp(arg, "--quiet"))
2273                DIFF_OPT_SET(options, QUIET);
2274        else if (!strcmp(arg, "--ext-diff"))
2275                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2276        else if (!strcmp(arg, "--no-ext-diff"))
2277                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2278
2279        /* misc options */
2280        else if (!strcmp(arg, "-z"))
2281                options->line_termination = 0;
2282        else if (!prefixcmp(arg, "-l"))
2283                options->rename_limit = strtoul(arg+2, NULL, 10);
2284        else if (!prefixcmp(arg, "-S"))
2285                options->pickaxe = arg + 2;
2286        else if (!strcmp(arg, "--pickaxe-all"))
2287                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2288        else if (!strcmp(arg, "--pickaxe-regex"))
2289                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2290        else if (!prefixcmp(arg, "-O"))
2291                options->orderfile = arg + 2;
2292        else if (!prefixcmp(arg, "--diff-filter="))
2293                options->filter = arg + 14;
2294        else if (!strcmp(arg, "--abbrev"))
2295                options->abbrev = DEFAULT_ABBREV;
2296        else if (!prefixcmp(arg, "--abbrev=")) {
2297                options->abbrev = strtoul(arg + 9, NULL, 10);
2298                if (options->abbrev < MINIMUM_ABBREV)
2299                        options->abbrev = MINIMUM_ABBREV;
2300                else if (40 < options->abbrev)
2301                        options->abbrev = 40;
2302        }
2303        else if (!prefixcmp(arg, "--src-prefix="))
2304                options->a_prefix = arg + 13;
2305        else if (!prefixcmp(arg, "--dst-prefix="))
2306                options->b_prefix = arg + 13;
2307        else if (!strcmp(arg, "--no-prefix"))
2308                options->a_prefix = options->b_prefix = "";
2309        else
2310                return 0;
2311        return 1;
2312}
2313
2314static int parse_num(const char **cp_p)
2315{
2316        unsigned long num, scale;
2317        int ch, dot;
2318        const char *cp = *cp_p;
2319
2320        num = 0;
2321        scale = 1;
2322        dot = 0;
2323        for(;;) {
2324                ch = *cp;
2325                if ( !dot && ch == '.' ) {
2326                        scale = 1;
2327                        dot = 1;
2328                } else if ( ch == '%' ) {
2329                        scale = dot ? scale*100 : 100;
2330                        cp++;   /* % is always at the end */
2331                        break;
2332                } else if ( ch >= '0' && ch <= '9' ) {
2333                        if ( scale < 100000 ) {
2334                                scale *= 10;
2335                                num = (num*10) + (ch-'0');
2336                        }
2337                } else {
2338                        break;
2339                }
2340                cp++;
2341        }
2342        *cp_p = cp;
2343
2344        /* user says num divided by scale and we say internally that
2345         * is MAX_SCORE * num / scale.
2346         */
2347        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2348}
2349
2350static int diff_scoreopt_parse(const char *opt)
2351{
2352        int opt1, opt2, cmd;
2353
2354        if (*opt++ != '-')
2355                return -1;
2356        cmd = *opt++;
2357        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2358                return -1; /* that is not a -M, -C nor -B option */
2359
2360        opt1 = parse_num(&opt);
2361        if (cmd != 'B')
2362                opt2 = 0;
2363        else {
2364                if (*opt == 0)
2365                        opt2 = 0;
2366                else if (*opt != '/')
2367                        return -1; /* we expect -B80/99 or -B80 */
2368                else {
2369                        opt++;
2370                        opt2 = parse_num(&opt);
2371                }
2372        }
2373        if (*opt != 0)
2374                return -1;
2375        return opt1 | (opt2 << 16);
2376}
2377
2378struct diff_queue_struct diff_queued_diff;
2379
2380void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2381{
2382        if (queue->alloc <= queue->nr) {
2383                queue->alloc = alloc_nr(queue->alloc);
2384                queue->queue = xrealloc(queue->queue,
2385                                        sizeof(dp) * queue->alloc);
2386        }
2387        queue->queue[queue->nr++] = dp;
2388}
2389
2390struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2391                                 struct diff_filespec *one,
2392                                 struct diff_filespec *two)
2393{
2394        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2395        dp->one = one;
2396        dp->two = two;
2397        if (queue)
2398                diff_q(queue, dp);
2399        return dp;
2400}
2401
2402void diff_free_filepair(struct diff_filepair *p)
2403{
2404        free_filespec(p->one);
2405        free_filespec(p->two);
2406        free(p);
2407}
2408
2409/* This is different from find_unique_abbrev() in that
2410 * it stuffs the result with dots for alignment.
2411 */
2412const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2413{
2414        int abblen;
2415        const char *abbrev;
2416        if (len == 40)
2417                return sha1_to_hex(sha1);
2418
2419        abbrev = find_unique_abbrev(sha1, len);
2420        if (!abbrev)
2421                return sha1_to_hex(sha1);
2422        abblen = strlen(abbrev);
2423        if (abblen < 37) {
2424                static char hex[41];
2425                if (len < abblen && abblen <= len + 2)
2426                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2427                else
2428                        sprintf(hex, "%s...", abbrev);
2429                return hex;
2430        }
2431        return sha1_to_hex(sha1);
2432}
2433
2434static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2435{
2436        int line_termination = opt->line_termination;
2437        int inter_name_termination = line_termination ? '\t' : '\0';
2438
2439        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2440                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2441                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2442                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2443        }
2444        if (p->score) {
2445                printf("%c%03d%c", p->status, similarity_index(p),
2446                           inter_name_termination);
2447        } else {
2448                printf("%c%c", p->status, inter_name_termination);
2449        }
2450
2451        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2452                write_name_quoted(p->one->path, stdout, inter_name_termination);
2453                write_name_quoted(p->two->path, stdout, line_termination);
2454        } else {
2455                const char *path = p->one->mode ? p->one->path : p->two->path;
2456                write_name_quoted(path, stdout, line_termination);
2457        }
2458}
2459
2460int diff_unmodified_pair(struct diff_filepair *p)
2461{
2462        /* This function is written stricter than necessary to support
2463         * the currently implemented transformers, but the idea is to
2464         * let transformers to produce diff_filepairs any way they want,
2465         * and filter and clean them up here before producing the output.
2466         */
2467        struct diff_filespec *one = p->one, *two = p->two;
2468
2469        if (DIFF_PAIR_UNMERGED(p))
2470                return 0; /* unmerged is interesting */
2471
2472        /* deletion, addition, mode or type change
2473         * and rename are all interesting.
2474         */
2475        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2476            DIFF_PAIR_MODE_CHANGED(p) ||
2477            strcmp(one->path, two->path))
2478                return 0;
2479
2480        /* both are valid and point at the same path.  that is, we are
2481         * dealing with a change.
2482         */
2483        if (one->sha1_valid && two->sha1_valid &&
2484            !hashcmp(one->sha1, two->sha1))
2485                return 1; /* no change */
2486        if (!one->sha1_valid && !two->sha1_valid)
2487                return 1; /* both look at the same file on the filesystem. */
2488        return 0;
2489}
2490
2491static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2492{
2493        if (diff_unmodified_pair(p))
2494                return;
2495
2496        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2497            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2498                return; /* no tree diffs in patch format */
2499
2500        run_diff(p, o);
2501}
2502
2503static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2504                            struct diffstat_t *diffstat)
2505{
2506        if (diff_unmodified_pair(p))
2507                return;
2508
2509        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2510            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2511                return; /* no tree diffs in patch format */
2512
2513        run_diffstat(p, o, diffstat);
2514}
2515
2516static void diff_flush_checkdiff(struct diff_filepair *p,
2517                struct diff_options *o)
2518{
2519        if (diff_unmodified_pair(p))
2520                return;
2521
2522        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2523            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2524                return; /* no tree diffs in patch format */
2525
2526        run_checkdiff(p, o);
2527}
2528
2529int diff_queue_is_empty(void)
2530{
2531        struct diff_queue_struct *q = &diff_queued_diff;
2532        int i;
2533        for (i = 0; i < q->nr; i++)
2534                if (!diff_unmodified_pair(q->queue[i]))
2535                        return 0;
2536        return 1;
2537}
2538
2539#if DIFF_DEBUG
2540void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2541{
2542        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2543                x, one ? one : "",
2544                s->path,
2545                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2546                s->mode,
2547                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2548        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2549                x, one ? one : "",
2550                s->size, s->xfrm_flags);
2551}
2552
2553void diff_debug_filepair(const struct diff_filepair *p, int i)
2554{
2555        diff_debug_filespec(p->one, i, "one");
2556        diff_debug_filespec(p->two, i, "two");
2557        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2558                p->score, p->status ? p->status : '?',
2559                p->one->rename_used, p->broken_pair);
2560}
2561
2562void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2563{
2564        int i;
2565        if (msg)
2566                fprintf(stderr, "%s\n", msg);
2567        fprintf(stderr, "q->nr = %d\n", q->nr);
2568        for (i = 0; i < q->nr; i++) {
2569                struct diff_filepair *p = q->queue[i];
2570                diff_debug_filepair(p, i);
2571        }
2572}
2573#endif
2574
2575static void diff_resolve_rename_copy(void)
2576{
2577        int i;
2578        struct diff_filepair *p;
2579        struct diff_queue_struct *q = &diff_queued_diff;
2580
2581        diff_debug_queue("resolve-rename-copy", q);
2582
2583        for (i = 0; i < q->nr; i++) {
2584                p = q->queue[i];
2585                p->status = 0; /* undecided */
2586                if (DIFF_PAIR_UNMERGED(p))
2587                        p->status = DIFF_STATUS_UNMERGED;
2588                else if (!DIFF_FILE_VALID(p->one))
2589                        p->status = DIFF_STATUS_ADDED;
2590                else if (!DIFF_FILE_VALID(p->two))
2591                        p->status = DIFF_STATUS_DELETED;
2592                else if (DIFF_PAIR_TYPE_CHANGED(p))
2593                        p->status = DIFF_STATUS_TYPE_CHANGED;
2594
2595                /* from this point on, we are dealing with a pair
2596                 * whose both sides are valid and of the same type, i.e.
2597                 * either in-place edit or rename/copy edit.
2598                 */
2599                else if (DIFF_PAIR_RENAME(p)) {
2600                        /*
2601                         * A rename might have re-connected a broken
2602                         * pair up, causing the pathnames to be the
2603                         * same again. If so, that's not a rename at
2604                         * all, just a modification..
2605                         *
2606                         * Otherwise, see if this source was used for
2607                         * multiple renames, in which case we decrement
2608                         * the count, and call it a copy.
2609                         */
2610                        if (!strcmp(p->one->path, p->two->path))
2611                                p->status = DIFF_STATUS_MODIFIED;
2612                        else if (--p->one->rename_used > 0)
2613                                p->status = DIFF_STATUS_COPIED;
2614                        else
2615                                p->status = DIFF_STATUS_RENAMED;
2616                }
2617                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2618                         p->one->mode != p->two->mode ||
2619                         is_null_sha1(p->one->sha1))
2620                        p->status = DIFF_STATUS_MODIFIED;
2621                else {
2622                        /* This is a "no-change" entry and should not
2623                         * happen anymore, but prepare for broken callers.
2624                         */
2625                        error("feeding unmodified %s to diffcore",
2626                              p->one->path);
2627                        p->status = DIFF_STATUS_UNKNOWN;
2628                }
2629        }
2630        diff_debug_queue("resolve-rename-copy done", q);
2631}
2632
2633static int check_pair_status(struct diff_filepair *p)
2634{
2635        switch (p->status) {
2636        case DIFF_STATUS_UNKNOWN:
2637                return 0;
2638        case 0:
2639                die("internal error in diff-resolve-rename-copy");
2640        default:
2641                return 1;
2642        }
2643}
2644
2645static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2646{
2647        int fmt = opt->output_format;
2648
2649        if (fmt & DIFF_FORMAT_CHECKDIFF)
2650                diff_flush_checkdiff(p, opt);
2651        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2652                diff_flush_raw(p, opt);
2653        else if (fmt & DIFF_FORMAT_NAME)
2654                write_name_quoted(p->two->path, stdout, opt->line_termination);
2655}
2656
2657static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2658{
2659        if (fs->mode)
2660                printf(" %s mode %06o ", newdelete, fs->mode);
2661        else
2662                printf(" %s ", newdelete);
2663        write_name_quoted(fs->path, stdout, '\n');
2664}
2665
2666
2667static void show_mode_change(struct diff_filepair *p, int show_name)
2668{
2669        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2670                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2671                        show_name ? ' ' : '\n');
2672                if (show_name) {
2673                        write_name_quoted(p->two->path, stdout, '\n');
2674                }
2675        }
2676}
2677
2678static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2679{
2680        char *names = pprint_rename(p->one->path, p->two->path);
2681
2682        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2683        free(names);
2684        show_mode_change(p, 0);
2685}
2686
2687static void diff_summary(struct diff_filepair *p)
2688{
2689        switch(p->status) {
2690        case DIFF_STATUS_DELETED:
2691                show_file_mode_name("delete", p->one);
2692                break;
2693        case DIFF_STATUS_ADDED:
2694                show_file_mode_name("create", p->two);
2695                break;
2696        case DIFF_STATUS_COPIED:
2697                show_rename_copy("copy", p);
2698                break;
2699        case DIFF_STATUS_RENAMED:
2700                show_rename_copy("rename", p);
2701                break;
2702        default:
2703                if (p->score) {
2704                        fputs(" rewrite ", stdout);
2705                        write_name_quoted(p->two->path, stdout, ' ');
2706                        printf("(%d%%)\n", similarity_index(p));
2707                }
2708                show_mode_change(p, !p->score);
2709                break;
2710        }
2711}
2712
2713struct patch_id_t {
2714        struct xdiff_emit_state xm;
2715        SHA_CTX *ctx;
2716        int patchlen;
2717};
2718
2719static int remove_space(char *line, int len)
2720{
2721        int i;
2722        char *dst = line;
2723        unsigned char c;
2724
2725        for (i = 0; i < len; i++)
2726                if (!isspace((c = line[i])))
2727                        *dst++ = c;
2728
2729        return dst - line;
2730}
2731
2732static void patch_id_consume(void *priv, char *line, unsigned long len)
2733{
2734        struct patch_id_t *data = priv;
2735        int new_len;
2736
2737        /* Ignore line numbers when computing the SHA1 of the patch */
2738        if (!prefixcmp(line, "@@ -"))
2739                return;
2740
2741        new_len = remove_space(line, len);
2742
2743        SHA1_Update(data->ctx, line, new_len);
2744        data->patchlen += new_len;
2745}
2746
2747/* returns 0 upon success, and writes result into sha1 */
2748static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2749{
2750        struct diff_queue_struct *q = &diff_queued_diff;
2751        int i;
2752        SHA_CTX ctx;
2753        struct patch_id_t data;
2754        char buffer[PATH_MAX * 4 + 20];
2755
2756        SHA1_Init(&ctx);
2757        memset(&data, 0, sizeof(struct patch_id_t));
2758        data.ctx = &ctx;
2759        data.xm.consume = patch_id_consume;
2760
2761        for (i = 0; i < q->nr; i++) {
2762                xpparam_t xpp;
2763                xdemitconf_t xecfg;
2764                xdemitcb_t ecb;
2765                mmfile_t mf1, mf2;
2766                struct diff_filepair *p = q->queue[i];
2767                int len1, len2;
2768
2769                memset(&xecfg, 0, sizeof(xecfg));
2770                if (p->status == 0)
2771                        return error("internal diff status error");
2772                if (p->status == DIFF_STATUS_UNKNOWN)
2773                        continue;
2774                if (diff_unmodified_pair(p))
2775                        continue;
2776                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2777                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2778                        continue;
2779                if (DIFF_PAIR_UNMERGED(p))
2780                        continue;
2781
2782                diff_fill_sha1_info(p->one);
2783                diff_fill_sha1_info(p->two);
2784                if (fill_mmfile(&mf1, p->one) < 0 ||
2785                                fill_mmfile(&mf2, p->two) < 0)
2786                        return error("unable to read files to diff");
2787
2788                len1 = remove_space(p->one->path, strlen(p->one->path));
2789                len2 = remove_space(p->two->path, strlen(p->two->path));
2790                if (p->one->mode == 0)
2791                        len1 = snprintf(buffer, sizeof(buffer),
2792                                        "diff--gita/%.*sb/%.*s"
2793                                        "newfilemode%06o"
2794                                        "---/dev/null"
2795                                        "+++b/%.*s",
2796                                        len1, p->one->path,
2797                                        len2, p->two->path,
2798                                        p->two->mode,
2799                                        len2, p->two->path);
2800                else if (p->two->mode == 0)
2801                        len1 = snprintf(buffer, sizeof(buffer),
2802                                        "diff--gita/%.*sb/%.*s"
2803                                        "deletedfilemode%06o"
2804                                        "---a/%.*s"
2805                                        "+++/dev/null",
2806                                        len1, p->one->path,
2807                                        len2, p->two->path,
2808                                        p->one->mode,
2809                                        len1, p->one->path);
2810                else
2811                        len1 = snprintf(buffer, sizeof(buffer),
2812                                        "diff--gita/%.*sb/%.*s"
2813                                        "---a/%.*s"
2814                                        "+++b/%.*s",
2815                                        len1, p->one->path,
2816                                        len2, p->two->path,
2817                                        len1, p->one->path,
2818                                        len2, p->two->path);
2819                SHA1_Update(&ctx, buffer, len1);
2820
2821                xpp.flags = XDF_NEED_MINIMAL;
2822                xecfg.ctxlen = 3;
2823                xecfg.flags = XDL_EMIT_FUNCNAMES;
2824                ecb.outf = xdiff_outf;
2825                ecb.priv = &data;
2826                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2827        }
2828
2829        SHA1_Final(sha1, &ctx);
2830        return 0;
2831}
2832
2833int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2834{
2835        struct diff_queue_struct *q = &diff_queued_diff;
2836        int i;
2837        int result = diff_get_patch_id(options, sha1);
2838
2839        for (i = 0; i < q->nr; i++)
2840                diff_free_filepair(q->queue[i]);
2841
2842        free(q->queue);
2843        q->queue = NULL;
2844        q->nr = q->alloc = 0;
2845
2846        return result;
2847}
2848
2849static int is_summary_empty(const struct diff_queue_struct *q)
2850{
2851        int i;
2852
2853        for (i = 0; i < q->nr; i++) {
2854                const struct diff_filepair *p = q->queue[i];
2855
2856                switch (p->status) {
2857                case DIFF_STATUS_DELETED:
2858                case DIFF_STATUS_ADDED:
2859                case DIFF_STATUS_COPIED:
2860                case DIFF_STATUS_RENAMED:
2861                        return 0;
2862                default:
2863                        if (p->score)
2864                                return 0;
2865                        if (p->one->mode && p->two->mode &&
2866                            p->one->mode != p->two->mode)
2867                                return 0;
2868                        break;
2869                }
2870        }
2871        return 1;
2872}
2873
2874void diff_flush(struct diff_options *options)
2875{
2876        struct diff_queue_struct *q = &diff_queued_diff;
2877        int i, output_format = options->output_format;
2878        int separator = 0;
2879
2880        /*
2881         * Order: raw, stat, summary, patch
2882         * or:    name/name-status/checkdiff (other bits clear)
2883         */
2884        if (!q->nr)
2885                goto free_queue;
2886
2887        if (output_format & (DIFF_FORMAT_RAW |
2888                             DIFF_FORMAT_NAME |
2889                             DIFF_FORMAT_NAME_STATUS |
2890                             DIFF_FORMAT_CHECKDIFF)) {
2891                for (i = 0; i < q->nr; i++) {
2892                        struct diff_filepair *p = q->queue[i];
2893                        if (check_pair_status(p))
2894                                flush_one_pair(p, options);
2895                }
2896                separator++;
2897        }
2898
2899        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2900                struct diffstat_t diffstat;
2901
2902                memset(&diffstat, 0, sizeof(struct diffstat_t));
2903                diffstat.xm.consume = diffstat_consume;
2904                for (i = 0; i < q->nr; i++) {
2905                        struct diff_filepair *p = q->queue[i];
2906                        if (check_pair_status(p))
2907                                diff_flush_stat(p, options, &diffstat);
2908                }
2909                if (output_format & DIFF_FORMAT_NUMSTAT)
2910                        show_numstat(&diffstat, options);
2911                if (output_format & DIFF_FORMAT_DIFFSTAT)
2912                        show_stats(&diffstat, options);
2913                if (output_format & DIFF_FORMAT_SHORTSTAT)
2914                        show_shortstats(&diffstat);
2915                free_diffstat_info(&diffstat);
2916                separator++;
2917        }
2918
2919        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2920                for (i = 0; i < q->nr; i++)
2921                        diff_summary(q->queue[i]);
2922                separator++;
2923        }
2924
2925        if (output_format & DIFF_FORMAT_PATCH) {
2926                if (separator) {
2927                        if (options->stat_sep) {
2928                                /* attach patch instead of inline */
2929                                fputs(options->stat_sep, stdout);
2930                        } else {
2931                                putchar(options->line_termination);
2932                        }
2933                }
2934
2935                for (i = 0; i < q->nr; i++) {
2936                        struct diff_filepair *p = q->queue[i];
2937                        if (check_pair_status(p))
2938                                diff_flush_patch(p, options);
2939                }
2940        }
2941
2942        if (output_format & DIFF_FORMAT_CALLBACK)
2943                options->format_callback(q, options, options->format_callback_data);
2944
2945        for (i = 0; i < q->nr; i++)
2946                diff_free_filepair(q->queue[i]);
2947free_queue:
2948        free(q->queue);
2949        q->queue = NULL;
2950        q->nr = q->alloc = 0;
2951}
2952
2953static void diffcore_apply_filter(const char *filter)
2954{
2955        int i;
2956        struct diff_queue_struct *q = &diff_queued_diff;
2957        struct diff_queue_struct outq;
2958        outq.queue = NULL;
2959        outq.nr = outq.alloc = 0;
2960
2961        if (!filter)
2962                return;
2963
2964        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2965                int found;
2966                for (i = found = 0; !found && i < q->nr; i++) {
2967                        struct diff_filepair *p = q->queue[i];
2968                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2969                             ((p->score &&
2970                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2971                              (!p->score &&
2972                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2973                            ((p->status != DIFF_STATUS_MODIFIED) &&
2974                             strchr(filter, p->status)))
2975                                found++;
2976                }
2977                if (found)
2978                        return;
2979
2980                /* otherwise we will clear the whole queue
2981                 * by copying the empty outq at the end of this
2982                 * function, but first clear the current entries
2983                 * in the queue.
2984                 */
2985                for (i = 0; i < q->nr; i++)
2986                        diff_free_filepair(q->queue[i]);
2987        }
2988        else {
2989                /* Only the matching ones */
2990                for (i = 0; i < q->nr; i++) {
2991                        struct diff_filepair *p = q->queue[i];
2992
2993                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2994                             ((p->score &&
2995                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2996                              (!p->score &&
2997                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2998                            ((p->status != DIFF_STATUS_MODIFIED) &&
2999                             strchr(filter, p->status)))
3000                                diff_q(&outq, p);
3001                        else
3002                                diff_free_filepair(p);
3003                }
3004        }
3005        free(q->queue);
3006        *q = outq;
3007}
3008
3009/* Check whether two filespecs with the same mode and size are identical */
3010static int diff_filespec_is_identical(struct diff_filespec *one,
3011                                      struct diff_filespec *two)
3012{
3013        if (S_ISGITLINK(one->mode)) {
3014                diff_fill_sha1_info(one);
3015                diff_fill_sha1_info(two);
3016                return !hashcmp(one->sha1, two->sha1);
3017        }
3018        if (diff_populate_filespec(one, 0))
3019                return 0;
3020        if (diff_populate_filespec(two, 0))
3021                return 0;
3022        return !memcmp(one->data, two->data, one->size);
3023}
3024
3025static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3026{
3027        int i;
3028        struct diff_queue_struct *q = &diff_queued_diff;
3029        struct diff_queue_struct outq;
3030        outq.queue = NULL;
3031        outq.nr = outq.alloc = 0;
3032
3033        for (i = 0; i < q->nr; i++) {
3034                struct diff_filepair *p = q->queue[i];
3035
3036                /*
3037                 * 1. Entries that come from stat info dirtyness
3038                 *    always have both sides (iow, not create/delete),
3039                 *    one side of the object name is unknown, with
3040                 *    the same mode and size.  Keep the ones that
3041                 *    do not match these criteria.  They have real
3042                 *    differences.
3043                 *
3044                 * 2. At this point, the file is known to be modified,
3045                 *    with the same mode and size, and the object
3046                 *    name of one side is unknown.  Need to inspect
3047                 *    the identical contents.
3048                 */
3049                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3050                    !DIFF_FILE_VALID(p->two) ||
3051                    (p->one->sha1_valid && p->two->sha1_valid) ||
3052                    (p->one->mode != p->two->mode) ||
3053                    diff_populate_filespec(p->one, 1) ||
3054                    diff_populate_filespec(p->two, 1) ||
3055                    (p->one->size != p->two->size) ||
3056                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3057                        diff_q(&outq, p);
3058                else {
3059                        /*
3060                         * The caller can subtract 1 from skip_stat_unmatch
3061                         * to determine how many paths were dirty only
3062                         * due to stat info mismatch.
3063                         */
3064                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3065                                diffopt->skip_stat_unmatch++;
3066                        diff_free_filepair(p);
3067                }
3068        }
3069        free(q->queue);
3070        *q = outq;
3071}
3072
3073void diffcore_std(struct diff_options *options)
3074{
3075        if (DIFF_OPT_TST(options, QUIET))
3076                return;
3077
3078        if (options->skip_stat_unmatch && !DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3079                diffcore_skip_stat_unmatch(options);
3080        if (options->break_opt != -1)
3081                diffcore_break(options->break_opt);
3082        if (options->detect_rename)
3083                diffcore_rename(options);
3084        if (options->break_opt != -1)
3085                diffcore_merge_broken();
3086        if (options->pickaxe)
3087                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3088        if (options->orderfile)
3089                diffcore_order(options->orderfile);
3090        diff_resolve_rename_copy();
3091        diffcore_apply_filter(options->filter);
3092
3093        if (diff_queued_diff.nr)
3094                DIFF_OPT_SET(options, HAS_CHANGES);
3095        else
3096                DIFF_OPT_CLR(options, HAS_CHANGES);
3097}
3098
3099int diff_result_code(struct diff_options *opt, int status)
3100{
3101        int result = 0;
3102        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3103            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3104                return status;
3105        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3106            DIFF_OPT_TST(opt, HAS_CHANGES))
3107                result |= 01;
3108        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3109            DIFF_OPT_TST(opt, CHECK_FAILED))
3110                result |= 02;
3111        return result;
3112}
3113
3114void diff_addremove(struct diff_options *options,
3115                    int addremove, unsigned mode,
3116                    const unsigned char *sha1,
3117                    const char *base, const char *path)
3118{
3119        char concatpath[PATH_MAX];
3120        struct diff_filespec *one, *two;
3121
3122        /* This may look odd, but it is a preparation for
3123         * feeding "there are unchanged files which should
3124         * not produce diffs, but when you are doing copy
3125         * detection you would need them, so here they are"
3126         * entries to the diff-core.  They will be prefixed
3127         * with something like '=' or '*' (I haven't decided
3128         * which but should not make any difference).
3129         * Feeding the same new and old to diff_change()
3130         * also has the same effect.
3131         * Before the final output happens, they are pruned after
3132         * merged into rename/copy pairs as appropriate.
3133         */
3134        if (DIFF_OPT_TST(options, REVERSE_DIFF))
3135                addremove = (addremove == '+' ? '-' :
3136                             addremove == '-' ? '+' : addremove);
3137
3138        if (!path) path = "";
3139        sprintf(concatpath, "%s%s", base, path);
3140        one = alloc_filespec(concatpath);
3141        two = alloc_filespec(concatpath);
3142
3143        if (addremove != '+')
3144                fill_filespec(one, sha1, mode);
3145        if (addremove != '-')
3146                fill_filespec(two, sha1, mode);
3147
3148        diff_queue(&diff_queued_diff, one, two);
3149        DIFF_OPT_SET(options, HAS_CHANGES);
3150}
3151
3152void diff_change(struct diff_options *options,
3153                 unsigned old_mode, unsigned new_mode,
3154                 const unsigned char *old_sha1,
3155                 const unsigned char *new_sha1,
3156                 const char *base, const char *path)
3157{
3158        char concatpath[PATH_MAX];
3159        struct diff_filespec *one, *two;
3160
3161        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3162                unsigned tmp;
3163                const unsigned char *tmp_c;
3164                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3165                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3166        }
3167        if (!path) path = "";
3168        sprintf(concatpath, "%s%s", base, path);
3169        one = alloc_filespec(concatpath);
3170        two = alloc_filespec(concatpath);
3171        fill_filespec(one, old_sha1, old_mode);
3172        fill_filespec(two, new_sha1, new_mode);
3173
3174        diff_queue(&diff_queued_diff, one, two);
3175        DIFF_OPT_SET(options, HAS_CHANGES);
3176}
3177
3178void diff_unmerge(struct diff_options *options,
3179                  const char *path,
3180                  unsigned mode, const unsigned char *sha1)
3181{
3182        struct diff_filespec *one, *two;
3183        one = alloc_filespec(path);
3184        two = alloc_filespec(path);
3185        fill_filespec(one, sha1, mode);
3186        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3187}