3e46ff8a75c5840a5f3374aa76e93cfb468c86f1
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13
  14#ifdef NO_FAST_WORKING_DIRECTORY
  15#define FAST_WORKING_DIRECTORY 0
  16#else
  17#define FAST_WORKING_DIRECTORY 1
  18#endif
  19
  20static int diff_detect_rename_default;
  21static int diff_rename_limit_default = 100;
  22static int diff_use_color_default;
  23int diff_auto_refresh_index = 1;
  24
  25static char diff_colors[][COLOR_MAXLEN] = {
  26        "\033[m",       /* reset */
  27        "",             /* PLAIN (normal) */
  28        "\033[1m",      /* METAINFO (bold) */
  29        "\033[36m",     /* FRAGINFO (cyan) */
  30        "\033[31m",     /* OLD (red) */
  31        "\033[32m",     /* NEW (green) */
  32        "\033[33m",     /* COMMIT (yellow) */
  33        "\033[41m",     /* WHITESPACE (red background) */
  34};
  35
  36static int parse_diff_color_slot(const char *var, int ofs)
  37{
  38        if (!strcasecmp(var+ofs, "plain"))
  39                return DIFF_PLAIN;
  40        if (!strcasecmp(var+ofs, "meta"))
  41                return DIFF_METAINFO;
  42        if (!strcasecmp(var+ofs, "frag"))
  43                return DIFF_FRAGINFO;
  44        if (!strcasecmp(var+ofs, "old"))
  45                return DIFF_FILE_OLD;
  46        if (!strcasecmp(var+ofs, "new"))
  47                return DIFF_FILE_NEW;
  48        if (!strcasecmp(var+ofs, "commit"))
  49                return DIFF_COMMIT;
  50        if (!strcasecmp(var+ofs, "whitespace"))
  51                return DIFF_WHITESPACE;
  52        die("bad config variable '%s'", var);
  53}
  54
  55static struct ll_diff_driver {
  56        const char *name;
  57        struct ll_diff_driver *next;
  58        char *cmd;
  59} *user_diff, **user_diff_tail;
  60
  61static void read_config_if_needed(void)
  62{
  63        if (!user_diff_tail) {
  64                user_diff_tail = &user_diff;
  65                git_config(git_diff_ui_config);
  66        }
  67}
  68
  69/*
  70 * Currently there is only "diff.<drivername>.command" variable;
  71 * because there are "diff.color.<slot>" variables, we are parsing
  72 * this in a bit convoluted way to allow low level diff driver
  73 * called "color".
  74 */
  75static int parse_lldiff_command(const char *var, const char *ep, const char *value)
  76{
  77        const char *name;
  78        int namelen;
  79        struct ll_diff_driver *drv;
  80
  81        name = var + 5;
  82        namelen = ep - name;
  83        for (drv = user_diff; drv; drv = drv->next)
  84                if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
  85                        break;
  86        if (!drv) {
  87                drv = xcalloc(1, sizeof(struct ll_diff_driver));
  88                drv->name = xmemdupz(name, namelen);
  89                if (!user_diff_tail)
  90                        user_diff_tail = &user_diff;
  91                *user_diff_tail = drv;
  92                user_diff_tail = &(drv->next);
  93        }
  94
  95        if (!value)
  96                return error("%s: lacks value", var);
  97        drv->cmd = strdup(value);
  98        return 0;
  99}
 100
 101/*
 102 * 'diff.<what>.funcname' attribute can be specified in the configuration
 103 * to define a customized regexp to find the beginning of a function to
 104 * be used for hunk header lines of "diff -p" style output.
 105 */
 106static struct funcname_pattern {
 107        char *name;
 108        char *pattern;
 109        struct funcname_pattern *next;
 110} *funcname_pattern_list;
 111
 112static int parse_funcname_pattern(const char *var, const char *ep, const char *value)
 113{
 114        const char *name;
 115        int namelen;
 116        struct funcname_pattern *pp;
 117
 118        name = var + 5; /* "diff." */
 119        namelen = ep - name;
 120
 121        for (pp = funcname_pattern_list; pp; pp = pp->next)
 122                if (!strncmp(pp->name, name, namelen) && !pp->name[namelen])
 123                        break;
 124        if (!pp) {
 125                pp = xcalloc(1, sizeof(*pp));
 126                pp->name = xmemdupz(name, namelen);
 127                pp->next = funcname_pattern_list;
 128                funcname_pattern_list = pp;
 129        }
 130        if (pp->pattern)
 131                free(pp->pattern);
 132        pp->pattern = xstrdup(value);
 133        return 0;
 134}
 135
 136/*
 137 * These are to give UI layer defaults.
 138 * The core-level commands such as git-diff-files should
 139 * never be affected by the setting of diff.renames
 140 * the user happens to have in the configuration file.
 141 */
 142int git_diff_ui_config(const char *var, const char *value)
 143{
 144        if (!strcmp(var, "diff.renamelimit")) {
 145                diff_rename_limit_default = git_config_int(var, value);
 146                return 0;
 147        }
 148        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 149                diff_use_color_default = git_config_colorbool(var, value, -1);
 150                return 0;
 151        }
 152        if (!strcmp(var, "diff.renames")) {
 153                if (!value)
 154                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 155                else if (!strcasecmp(value, "copies") ||
 156                         !strcasecmp(value, "copy"))
 157                        diff_detect_rename_default = DIFF_DETECT_COPY;
 158                else if (git_config_bool(var,value))
 159                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 160                return 0;
 161        }
 162        if (!strcmp(var, "diff.autorefreshindex")) {
 163                diff_auto_refresh_index = git_config_bool(var, value);
 164                return 0;
 165        }
 166        if (!prefixcmp(var, "diff.")) {
 167                const char *ep = strrchr(var, '.');
 168
 169                if (ep != var + 4) {
 170                        if (!strcmp(ep, ".command"))
 171                                return parse_lldiff_command(var, ep, value);
 172                        if (!strcmp(ep, ".funcname"))
 173                                return parse_funcname_pattern(var, ep, value);
 174                }
 175        }
 176        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 177                int slot = parse_diff_color_slot(var, 11);
 178                color_parse(value, var, diff_colors[slot]);
 179                return 0;
 180        }
 181
 182        return git_default_config(var, value);
 183}
 184
 185static char *quote_two(const char *one, const char *two)
 186{
 187        int need_one = quote_c_style(one, NULL, NULL, 1);
 188        int need_two = quote_c_style(two, NULL, NULL, 1);
 189        struct strbuf res;
 190
 191        strbuf_init(&res, 0);
 192        if (need_one + need_two) {
 193                strbuf_addch(&res, '"');
 194                quote_c_style(one, &res, NULL, 1);
 195                quote_c_style(two, &res, NULL, 1);
 196                strbuf_addch(&res, '"');
 197        } else {
 198                strbuf_addstr(&res, one);
 199                strbuf_addstr(&res, two);
 200        }
 201        return strbuf_detach(&res, NULL);
 202}
 203
 204static const char *external_diff(void)
 205{
 206        static const char *external_diff_cmd = NULL;
 207        static int done_preparing = 0;
 208
 209        if (done_preparing)
 210                return external_diff_cmd;
 211        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 212        done_preparing = 1;
 213        return external_diff_cmd;
 214}
 215
 216static struct diff_tempfile {
 217        const char *name; /* filename external diff should read from */
 218        char hex[41];
 219        char mode[10];
 220        char tmp_path[PATH_MAX];
 221} diff_temp[2];
 222
 223static int count_lines(const char *data, int size)
 224{
 225        int count, ch, completely_empty = 1, nl_just_seen = 0;
 226        count = 0;
 227        while (0 < size--) {
 228                ch = *data++;
 229                if (ch == '\n') {
 230                        count++;
 231                        nl_just_seen = 1;
 232                        completely_empty = 0;
 233                }
 234                else {
 235                        nl_just_seen = 0;
 236                        completely_empty = 0;
 237                }
 238        }
 239        if (completely_empty)
 240                return 0;
 241        if (!nl_just_seen)
 242                count++; /* no trailing newline */
 243        return count;
 244}
 245
 246static void print_line_count(int count)
 247{
 248        switch (count) {
 249        case 0:
 250                printf("0,0");
 251                break;
 252        case 1:
 253                printf("1");
 254                break;
 255        default:
 256                printf("1,%d", count);
 257                break;
 258        }
 259}
 260
 261static void copy_file(int prefix, const char *data, int size,
 262                const char *set, const char *reset)
 263{
 264        int ch, nl_just_seen = 1;
 265        while (0 < size--) {
 266                ch = *data++;
 267                if (nl_just_seen) {
 268                        fputs(set, stdout);
 269                        putchar(prefix);
 270                }
 271                if (ch == '\n') {
 272                        nl_just_seen = 1;
 273                        fputs(reset, stdout);
 274                } else
 275                        nl_just_seen = 0;
 276                putchar(ch);
 277        }
 278        if (!nl_just_seen)
 279                printf("%s\n\\ No newline at end of file\n", reset);
 280}
 281
 282static void emit_rewrite_diff(const char *name_a,
 283                              const char *name_b,
 284                              struct diff_filespec *one,
 285                              struct diff_filespec *two,
 286                              int color_diff)
 287{
 288        int lc_a, lc_b;
 289        const char *name_a_tab, *name_b_tab;
 290        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 291        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 292        const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
 293        const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
 294        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 295
 296        name_a += (*name_a == '/');
 297        name_b += (*name_b == '/');
 298        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 299        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 300
 301        diff_populate_filespec(one, 0);
 302        diff_populate_filespec(two, 0);
 303        lc_a = count_lines(one->data, one->size);
 304        lc_b = count_lines(two->data, two->size);
 305        printf("%s--- a/%s%s%s\n%s+++ b/%s%s%s\n%s@@ -",
 306               metainfo, name_a, name_a_tab, reset,
 307               metainfo, name_b, name_b_tab, reset, fraginfo);
 308        print_line_count(lc_a);
 309        printf(" +");
 310        print_line_count(lc_b);
 311        printf(" @@%s\n", reset);
 312        if (lc_a)
 313                copy_file('-', one->data, one->size, old, reset);
 314        if (lc_b)
 315                copy_file('+', two->data, two->size, new, reset);
 316}
 317
 318static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 319{
 320        if (!DIFF_FILE_VALID(one)) {
 321                mf->ptr = (char *)""; /* does not matter */
 322                mf->size = 0;
 323                return 0;
 324        }
 325        else if (diff_populate_filespec(one, 0))
 326                return -1;
 327        mf->ptr = one->data;
 328        mf->size = one->size;
 329        return 0;
 330}
 331
 332struct diff_words_buffer {
 333        mmfile_t text;
 334        long alloc;
 335        long current; /* output pointer */
 336        int suppressed_newline;
 337};
 338
 339static void diff_words_append(char *line, unsigned long len,
 340                struct diff_words_buffer *buffer)
 341{
 342        if (buffer->text.size + len > buffer->alloc) {
 343                buffer->alloc = (buffer->text.size + len) * 3 / 2;
 344                buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
 345        }
 346        line++;
 347        len--;
 348        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 349        buffer->text.size += len;
 350}
 351
 352struct diff_words_data {
 353        struct xdiff_emit_state xm;
 354        struct diff_words_buffer minus, plus;
 355};
 356
 357static void print_word(struct diff_words_buffer *buffer, int len, int color,
 358                int suppress_newline)
 359{
 360        const char *ptr;
 361        int eol = 0;
 362
 363        if (len == 0)
 364                return;
 365
 366        ptr  = buffer->text.ptr + buffer->current;
 367        buffer->current += len;
 368
 369        if (ptr[len - 1] == '\n') {
 370                eol = 1;
 371                len--;
 372        }
 373
 374        fputs(diff_get_color(1, color), stdout);
 375        fwrite(ptr, len, 1, stdout);
 376        fputs(diff_get_color(1, DIFF_RESET), stdout);
 377
 378        if (eol) {
 379                if (suppress_newline)
 380                        buffer->suppressed_newline = 1;
 381                else
 382                        putchar('\n');
 383        }
 384}
 385
 386static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 387{
 388        struct diff_words_data *diff_words = priv;
 389
 390        if (diff_words->minus.suppressed_newline) {
 391                if (line[0] != '+')
 392                        putchar('\n');
 393                diff_words->minus.suppressed_newline = 0;
 394        }
 395
 396        len--;
 397        switch (line[0]) {
 398                case '-':
 399                        print_word(&diff_words->minus, len, DIFF_FILE_OLD, 1);
 400                        break;
 401                case '+':
 402                        print_word(&diff_words->plus, len, DIFF_FILE_NEW, 0);
 403                        break;
 404                case ' ':
 405                        print_word(&diff_words->plus, len, DIFF_PLAIN, 0);
 406                        diff_words->minus.current += len;
 407                        break;
 408        }
 409}
 410
 411/* this executes the word diff on the accumulated buffers */
 412static void diff_words_show(struct diff_words_data *diff_words)
 413{
 414        xpparam_t xpp;
 415        xdemitconf_t xecfg;
 416        xdemitcb_t ecb;
 417        mmfile_t minus, plus;
 418        int i;
 419
 420        memset(&xecfg, 0, sizeof(xecfg));
 421        minus.size = diff_words->minus.text.size;
 422        minus.ptr = xmalloc(minus.size);
 423        memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
 424        for (i = 0; i < minus.size; i++)
 425                if (isspace(minus.ptr[i]))
 426                        minus.ptr[i] = '\n';
 427        diff_words->minus.current = 0;
 428
 429        plus.size = diff_words->plus.text.size;
 430        plus.ptr = xmalloc(plus.size);
 431        memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
 432        for (i = 0; i < plus.size; i++)
 433                if (isspace(plus.ptr[i]))
 434                        plus.ptr[i] = '\n';
 435        diff_words->plus.current = 0;
 436
 437        xpp.flags = XDF_NEED_MINIMAL;
 438        xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
 439        ecb.outf = xdiff_outf;
 440        ecb.priv = diff_words;
 441        diff_words->xm.consume = fn_out_diff_words_aux;
 442        xdi_diff(&minus, &plus, &xpp, &xecfg, &ecb);
 443
 444        free(minus.ptr);
 445        free(plus.ptr);
 446        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 447
 448        if (diff_words->minus.suppressed_newline) {
 449                putchar('\n');
 450                diff_words->minus.suppressed_newline = 0;
 451        }
 452}
 453
 454struct emit_callback {
 455        struct xdiff_emit_state xm;
 456        int nparents, color_diff;
 457        unsigned ws_rule;
 458        const char **label_path;
 459        struct diff_words_data *diff_words;
 460        int *found_changesp;
 461};
 462
 463static void free_diff_words_data(struct emit_callback *ecbdata)
 464{
 465        if (ecbdata->diff_words) {
 466                /* flush buffers */
 467                if (ecbdata->diff_words->minus.text.size ||
 468                                ecbdata->diff_words->plus.text.size)
 469                        diff_words_show(ecbdata->diff_words);
 470
 471                if (ecbdata->diff_words->minus.text.ptr)
 472                        free (ecbdata->diff_words->minus.text.ptr);
 473                if (ecbdata->diff_words->plus.text.ptr)
 474                        free (ecbdata->diff_words->plus.text.ptr);
 475                free(ecbdata->diff_words);
 476                ecbdata->diff_words = NULL;
 477        }
 478}
 479
 480const char *diff_get_color(int diff_use_color, enum color_diff ix)
 481{
 482        if (diff_use_color)
 483                return diff_colors[ix];
 484        return "";
 485}
 486
 487static void emit_line(const char *set, const char *reset, const char *line, int len)
 488{
 489        if (len > 0 && line[len-1] == '\n')
 490                len--;
 491        fputs(set, stdout);
 492        fwrite(line, len, 1, stdout);
 493        puts(reset);
 494}
 495
 496static void emit_line_with_ws(int nparents,
 497                              const char *set, const char *reset, const char *ws,
 498                              const char *line, int len, unsigned ws_rule)
 499{
 500        int col0 = nparents;
 501        int last_tab_in_indent = -1;
 502        int last_space_in_indent = -1;
 503        int i;
 504        int tail = len;
 505        int need_highlight_leading_space = 0;
 506        /*
 507         * The line is a newly added line.  Does it have funny leading
 508         * whitespaces?  In indent, SP should never precede a TAB.  In
 509         * addition, under "indent with non tab" rule, there should not
 510         * be more than 8 consecutive spaces.
 511         */
 512        for (i = col0; i < len; i++) {
 513                if (line[i] == '\t') {
 514                        last_tab_in_indent = i;
 515                        if ((ws_rule & WS_SPACE_BEFORE_TAB) &&
 516                            0 <= last_space_in_indent)
 517                                need_highlight_leading_space = 1;
 518                }
 519                else if (line[i] == ' ')
 520                        last_space_in_indent = i;
 521                else
 522                        break;
 523        }
 524        if ((ws_rule & WS_INDENT_WITH_NON_TAB) &&
 525            0 <= last_space_in_indent &&
 526            last_tab_in_indent < 0 &&
 527            8 <= (i - col0)) {
 528                last_tab_in_indent = i;
 529                need_highlight_leading_space = 1;
 530        }
 531        fputs(set, stdout);
 532        fwrite(line, col0, 1, stdout);
 533        fputs(reset, stdout);
 534        if (((i == len) || line[i] == '\n') && i != col0) {
 535                /* The whole line was indent */
 536                emit_line(ws, reset, line + col0, len - col0);
 537                return;
 538        }
 539        i = col0;
 540        if (need_highlight_leading_space) {
 541                while (i < last_tab_in_indent) {
 542                        if (line[i] == ' ') {
 543                                fputs(ws, stdout);
 544                                putchar(' ');
 545                                fputs(reset, stdout);
 546                        }
 547                        else
 548                                putchar(line[i]);
 549                        i++;
 550                }
 551        }
 552        tail = len - 1;
 553        if (line[tail] == '\n' && i < tail)
 554                tail--;
 555        if (ws_rule & WS_TRAILING_SPACE) {
 556                while (i < tail) {
 557                        if (!isspace(line[tail]))
 558                                break;
 559                        tail--;
 560                }
 561        }
 562        if ((i < tail && line[tail + 1] != '\n')) {
 563                /* This has whitespace between tail+1..len */
 564                fputs(set, stdout);
 565                fwrite(line + i, tail - i + 1, 1, stdout);
 566                fputs(reset, stdout);
 567                emit_line(ws, reset, line + tail + 1, len - tail - 1);
 568        }
 569        else
 570                emit_line(set, reset, line + i, len - i);
 571}
 572
 573static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 574{
 575        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 576        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 577
 578        if (!*ws)
 579                emit_line(set, reset, line, len);
 580        else
 581                emit_line_with_ws(ecbdata->nparents, set, reset, ws,
 582                                  line, len, ecbdata->ws_rule);
 583}
 584
 585static void fn_out_consume(void *priv, char *line, unsigned long len)
 586{
 587        int i;
 588        int color;
 589        struct emit_callback *ecbdata = priv;
 590        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 591        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 592
 593        *(ecbdata->found_changesp) = 1;
 594
 595        if (ecbdata->label_path[0]) {
 596                const char *name_a_tab, *name_b_tab;
 597
 598                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 599                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 600
 601                printf("%s--- %s%s%s\n",
 602                       set, ecbdata->label_path[0], reset, name_a_tab);
 603                printf("%s+++ %s%s%s\n",
 604                       set, ecbdata->label_path[1], reset, name_b_tab);
 605                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 606        }
 607
 608        /* This is not really necessary for now because
 609         * this codepath only deals with two-way diffs.
 610         */
 611        for (i = 0; i < len && line[i] == '@'; i++)
 612                ;
 613        if (2 <= i && i < len && line[i] == ' ') {
 614                ecbdata->nparents = i - 1;
 615                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 616                          reset, line, len);
 617                return;
 618        }
 619
 620        if (len < ecbdata->nparents) {
 621                set = reset;
 622                emit_line(reset, reset, line, len);
 623                return;
 624        }
 625
 626        color = DIFF_PLAIN;
 627        if (ecbdata->diff_words && ecbdata->nparents != 1)
 628                /* fall back to normal diff */
 629                free_diff_words_data(ecbdata);
 630        if (ecbdata->diff_words) {
 631                if (line[0] == '-') {
 632                        diff_words_append(line, len,
 633                                          &ecbdata->diff_words->minus);
 634                        return;
 635                } else if (line[0] == '+') {
 636                        diff_words_append(line, len,
 637                                          &ecbdata->diff_words->plus);
 638                        return;
 639                }
 640                if (ecbdata->diff_words->minus.text.size ||
 641                    ecbdata->diff_words->plus.text.size)
 642                        diff_words_show(ecbdata->diff_words);
 643                line++;
 644                len--;
 645                emit_line(set, reset, line, len);
 646                return;
 647        }
 648        for (i = 0; i < ecbdata->nparents && len; i++) {
 649                if (line[i] == '-')
 650                        color = DIFF_FILE_OLD;
 651                else if (line[i] == '+')
 652                        color = DIFF_FILE_NEW;
 653        }
 654
 655        if (color != DIFF_FILE_NEW) {
 656                emit_line(diff_get_color(ecbdata->color_diff, color),
 657                          reset, line, len);
 658                return;
 659        }
 660        emit_add_line(reset, ecbdata, line, len);
 661}
 662
 663static char *pprint_rename(const char *a, const char *b)
 664{
 665        const char *old = a;
 666        const char *new = b;
 667        struct strbuf name;
 668        int pfx_length, sfx_length;
 669        int len_a = strlen(a);
 670        int len_b = strlen(b);
 671        int a_midlen, b_midlen;
 672        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 673        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 674
 675        strbuf_init(&name, 0);
 676        if (qlen_a || qlen_b) {
 677                quote_c_style(a, &name, NULL, 0);
 678                strbuf_addstr(&name, " => ");
 679                quote_c_style(b, &name, NULL, 0);
 680                return strbuf_detach(&name, NULL);
 681        }
 682
 683        /* Find common prefix */
 684        pfx_length = 0;
 685        while (*old && *new && *old == *new) {
 686                if (*old == '/')
 687                        pfx_length = old - a + 1;
 688                old++;
 689                new++;
 690        }
 691
 692        /* Find common suffix */
 693        old = a + len_a;
 694        new = b + len_b;
 695        sfx_length = 0;
 696        while (a <= old && b <= new && *old == *new) {
 697                if (*old == '/')
 698                        sfx_length = len_a - (old - a);
 699                old--;
 700                new--;
 701        }
 702
 703        /*
 704         * pfx{mid-a => mid-b}sfx
 705         * {pfx-a => pfx-b}sfx
 706         * pfx{sfx-a => sfx-b}
 707         * name-a => name-b
 708         */
 709        a_midlen = len_a - pfx_length - sfx_length;
 710        b_midlen = len_b - pfx_length - sfx_length;
 711        if (a_midlen < 0)
 712                a_midlen = 0;
 713        if (b_midlen < 0)
 714                b_midlen = 0;
 715
 716        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 717        if (pfx_length + sfx_length) {
 718                strbuf_add(&name, a, pfx_length);
 719                strbuf_addch(&name, '{');
 720        }
 721        strbuf_add(&name, a + pfx_length, a_midlen);
 722        strbuf_addstr(&name, " => ");
 723        strbuf_add(&name, b + pfx_length, b_midlen);
 724        if (pfx_length + sfx_length) {
 725                strbuf_addch(&name, '}');
 726                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 727        }
 728        return strbuf_detach(&name, NULL);
 729}
 730
 731struct diffstat_t {
 732        struct xdiff_emit_state xm;
 733
 734        int nr;
 735        int alloc;
 736        struct diffstat_file {
 737                char *from_name;
 738                char *name;
 739                char *print_name;
 740                unsigned is_unmerged:1;
 741                unsigned is_binary:1;
 742                unsigned is_renamed:1;
 743                unsigned int added, deleted;
 744        } **files;
 745};
 746
 747static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 748                                          const char *name_a,
 749                                          const char *name_b)
 750{
 751        struct diffstat_file *x;
 752        x = xcalloc(sizeof (*x), 1);
 753        if (diffstat->nr == diffstat->alloc) {
 754                diffstat->alloc = alloc_nr(diffstat->alloc);
 755                diffstat->files = xrealloc(diffstat->files,
 756                                diffstat->alloc * sizeof(x));
 757        }
 758        diffstat->files[diffstat->nr++] = x;
 759        if (name_b) {
 760                x->from_name = xstrdup(name_a);
 761                x->name = xstrdup(name_b);
 762                x->is_renamed = 1;
 763        }
 764        else {
 765                x->from_name = NULL;
 766                x->name = xstrdup(name_a);
 767        }
 768        return x;
 769}
 770
 771static void diffstat_consume(void *priv, char *line, unsigned long len)
 772{
 773        struct diffstat_t *diffstat = priv;
 774        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 775
 776        if (line[0] == '+')
 777                x->added++;
 778        else if (line[0] == '-')
 779                x->deleted++;
 780}
 781
 782const char mime_boundary_leader[] = "------------";
 783
 784static int scale_linear(int it, int width, int max_change)
 785{
 786        /*
 787         * make sure that at least one '-' is printed if there were deletions,
 788         * and likewise for '+'.
 789         */
 790        if (max_change < 2)
 791                return it;
 792        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 793}
 794
 795static void show_name(const char *prefix, const char *name, int len,
 796                      const char *reset, const char *set)
 797{
 798        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 799}
 800
 801static void show_graph(char ch, int cnt, const char *set, const char *reset)
 802{
 803        if (cnt <= 0)
 804                return;
 805        printf("%s", set);
 806        while (cnt--)
 807                putchar(ch);
 808        printf("%s", reset);
 809}
 810
 811static void fill_print_name(struct diffstat_file *file)
 812{
 813        char *pname;
 814
 815        if (file->print_name)
 816                return;
 817
 818        if (!file->is_renamed) {
 819                struct strbuf buf;
 820                strbuf_init(&buf, 0);
 821                if (quote_c_style(file->name, &buf, NULL, 0)) {
 822                        pname = strbuf_detach(&buf, NULL);
 823                } else {
 824                        pname = file->name;
 825                        strbuf_release(&buf);
 826                }
 827        } else {
 828                pname = pprint_rename(file->from_name, file->name);
 829        }
 830        file->print_name = pname;
 831}
 832
 833static void show_stats(struct diffstat_t* data, struct diff_options *options)
 834{
 835        int i, len, add, del, total, adds = 0, dels = 0;
 836        int max_change = 0, max_len = 0;
 837        int total_files = data->nr;
 838        int width, name_width;
 839        const char *reset, *set, *add_c, *del_c;
 840
 841        if (data->nr == 0)
 842                return;
 843
 844        width = options->stat_width ? options->stat_width : 80;
 845        name_width = options->stat_name_width ? options->stat_name_width : 50;
 846
 847        /* Sanity: give at least 5 columns to the graph,
 848         * but leave at least 10 columns for the name.
 849         */
 850        if (width < name_width + 15) {
 851                if (name_width <= 25)
 852                        width = name_width + 15;
 853                else
 854                        name_width = width - 15;
 855        }
 856
 857        /* Find the longest filename and max number of changes */
 858        reset = diff_get_color_opt(options, DIFF_RESET);
 859        set   = diff_get_color_opt(options, DIFF_PLAIN);
 860        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
 861        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
 862
 863        for (i = 0; i < data->nr; i++) {
 864                struct diffstat_file *file = data->files[i];
 865                int change = file->added + file->deleted;
 866                fill_print_name(file);
 867                len = strlen(file->print_name);
 868                if (max_len < len)
 869                        max_len = len;
 870
 871                if (file->is_binary || file->is_unmerged)
 872                        continue;
 873                if (max_change < change)
 874                        max_change = change;
 875        }
 876
 877        /* Compute the width of the graph part;
 878         * 10 is for one blank at the beginning of the line plus
 879         * " | count " between the name and the graph.
 880         *
 881         * From here on, name_width is the width of the name area,
 882         * and width is the width of the graph area.
 883         */
 884        name_width = (name_width < max_len) ? name_width : max_len;
 885        if (width < (name_width + 10) + max_change)
 886                width = width - (name_width + 10);
 887        else
 888                width = max_change;
 889
 890        for (i = 0; i < data->nr; i++) {
 891                const char *prefix = "";
 892                char *name = data->files[i]->print_name;
 893                int added = data->files[i]->added;
 894                int deleted = data->files[i]->deleted;
 895                int name_len;
 896
 897                /*
 898                 * "scale" the filename
 899                 */
 900                len = name_width;
 901                name_len = strlen(name);
 902                if (name_width < name_len) {
 903                        char *slash;
 904                        prefix = "...";
 905                        len -= 3;
 906                        name += name_len - len;
 907                        slash = strchr(name, '/');
 908                        if (slash)
 909                                name = slash;
 910                }
 911
 912                if (data->files[i]->is_binary) {
 913                        show_name(prefix, name, len, reset, set);
 914                        printf("  Bin ");
 915                        printf("%s%d%s", del_c, deleted, reset);
 916                        printf(" -> ");
 917                        printf("%s%d%s", add_c, added, reset);
 918                        printf(" bytes");
 919                        printf("\n");
 920                        continue;
 921                }
 922                else if (data->files[i]->is_unmerged) {
 923                        show_name(prefix, name, len, reset, set);
 924                        printf("  Unmerged\n");
 925                        continue;
 926                }
 927                else if (!data->files[i]->is_renamed &&
 928                         (added + deleted == 0)) {
 929                        total_files--;
 930                        continue;
 931                }
 932
 933                /*
 934                 * scale the add/delete
 935                 */
 936                add = added;
 937                del = deleted;
 938                total = add + del;
 939                adds += add;
 940                dels += del;
 941
 942                if (width <= max_change) {
 943                        add = scale_linear(add, width, max_change);
 944                        del = scale_linear(del, width, max_change);
 945                        total = add + del;
 946                }
 947                show_name(prefix, name, len, reset, set);
 948                printf("%5d ", added + deleted);
 949                show_graph('+', add, add_c, reset);
 950                show_graph('-', del, del_c, reset);
 951                putchar('\n');
 952        }
 953        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 954               set, total_files, adds, dels, reset);
 955}
 956
 957static void show_shortstats(struct diffstat_t* data)
 958{
 959        int i, adds = 0, dels = 0, total_files = data->nr;
 960
 961        if (data->nr == 0)
 962                return;
 963
 964        for (i = 0; i < data->nr; i++) {
 965                if (!data->files[i]->is_binary &&
 966                    !data->files[i]->is_unmerged) {
 967                        int added = data->files[i]->added;
 968                        int deleted= data->files[i]->deleted;
 969                        if (!data->files[i]->is_renamed &&
 970                            (added + deleted == 0)) {
 971                                total_files--;
 972                        } else {
 973                                adds += added;
 974                                dels += deleted;
 975                        }
 976                }
 977        }
 978        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 979               total_files, adds, dels);
 980}
 981
 982static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 983{
 984        int i;
 985
 986        if (data->nr == 0)
 987                return;
 988
 989        for (i = 0; i < data->nr; i++) {
 990                struct diffstat_file *file = data->files[i];
 991
 992                if (file->is_binary)
 993                        printf("-\t-\t");
 994                else
 995                        printf("%d\t%d\t", file->added, file->deleted);
 996                if (options->line_termination) {
 997                        fill_print_name(file);
 998                        if (!file->is_renamed)
 999                                write_name_quoted(file->name, stdout,
1000                                                  options->line_termination);
1001                        else {
1002                                fputs(file->print_name, stdout);
1003                                putchar(options->line_termination);
1004                        }
1005                } else {
1006                        if (file->is_renamed) {
1007                                putchar('\0');
1008                                write_name_quoted(file->from_name, stdout, '\0');
1009                        }
1010                        write_name_quoted(file->name, stdout, '\0');
1011                }
1012        }
1013}
1014
1015static void free_diffstat_info(struct diffstat_t *diffstat)
1016{
1017        int i;
1018        for (i = 0; i < diffstat->nr; i++) {
1019                struct diffstat_file *f = diffstat->files[i];
1020                if (f->name != f->print_name)
1021                        free(f->print_name);
1022                free(f->name);
1023                free(f->from_name);
1024                free(f);
1025        }
1026        free(diffstat->files);
1027}
1028
1029struct checkdiff_t {
1030        struct xdiff_emit_state xm;
1031        const char *filename;
1032        int lineno, color_diff;
1033        unsigned ws_rule;
1034        unsigned status;
1035};
1036
1037static void checkdiff_consume(void *priv, char *line, unsigned long len)
1038{
1039        struct checkdiff_t *data = priv;
1040        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
1041        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
1042        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
1043
1044        if (line[0] == '+') {
1045                int i, spaces = 0, space_before_tab = 0, white_space_at_end = 0;
1046
1047                /* check space before tab */
1048                for (i = 1; i < len; i++) {
1049                        if (line[i] == ' ')
1050                                spaces++;
1051                        else if (line[i] == '\t') {
1052                                if (spaces) {
1053                                        space_before_tab = 1;
1054                                        break;
1055                                }
1056                        }
1057                        else
1058                                break;
1059                }
1060
1061                /* check whitespace at line end */
1062                if (line[len - 1] == '\n')
1063                        len--;
1064                if (isspace(line[len - 1]))
1065                        white_space_at_end = 1;
1066
1067                if (space_before_tab || white_space_at_end) {
1068                        data->status = 1;
1069                        printf("%s:%d: %s", data->filename, data->lineno, ws);
1070                        if (space_before_tab) {
1071                                printf("space before tab");
1072                                if (white_space_at_end)
1073                                        putchar(',');
1074                        }
1075                        if (white_space_at_end)
1076                                printf("whitespace at end");
1077                        printf(":%s ", reset);
1078                        emit_line_with_ws(1, set, reset, ws, line, len,
1079                                          data->ws_rule);
1080                }
1081
1082                data->lineno++;
1083        } else if (line[0] == ' ')
1084                data->lineno++;
1085        else if (line[0] == '@') {
1086                char *plus = strchr(line, '+');
1087                if (plus)
1088                        data->lineno = strtol(plus, NULL, 10);
1089                else
1090                        die("invalid diff");
1091        }
1092}
1093
1094static unsigned char *deflate_it(char *data,
1095                                 unsigned long size,
1096                                 unsigned long *result_size)
1097{
1098        int bound;
1099        unsigned char *deflated;
1100        z_stream stream;
1101
1102        memset(&stream, 0, sizeof(stream));
1103        deflateInit(&stream, zlib_compression_level);
1104        bound = deflateBound(&stream, size);
1105        deflated = xmalloc(bound);
1106        stream.next_out = deflated;
1107        stream.avail_out = bound;
1108
1109        stream.next_in = (unsigned char *)data;
1110        stream.avail_in = size;
1111        while (deflate(&stream, Z_FINISH) == Z_OK)
1112                ; /* nothing */
1113        deflateEnd(&stream);
1114        *result_size = stream.total_out;
1115        return deflated;
1116}
1117
1118static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1119{
1120        void *cp;
1121        void *delta;
1122        void *deflated;
1123        void *data;
1124        unsigned long orig_size;
1125        unsigned long delta_size;
1126        unsigned long deflate_size;
1127        unsigned long data_size;
1128
1129        /* We could do deflated delta, or we could do just deflated two,
1130         * whichever is smaller.
1131         */
1132        delta = NULL;
1133        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1134        if (one->size && two->size) {
1135                delta = diff_delta(one->ptr, one->size,
1136                                   two->ptr, two->size,
1137                                   &delta_size, deflate_size);
1138                if (delta) {
1139                        void *to_free = delta;
1140                        orig_size = delta_size;
1141                        delta = deflate_it(delta, delta_size, &delta_size);
1142                        free(to_free);
1143                }
1144        }
1145
1146        if (delta && delta_size < deflate_size) {
1147                printf("delta %lu\n", orig_size);
1148                free(deflated);
1149                data = delta;
1150                data_size = delta_size;
1151        }
1152        else {
1153                printf("literal %lu\n", two->size);
1154                free(delta);
1155                data = deflated;
1156                data_size = deflate_size;
1157        }
1158
1159        /* emit data encoded in base85 */
1160        cp = data;
1161        while (data_size) {
1162                int bytes = (52 < data_size) ? 52 : data_size;
1163                char line[70];
1164                data_size -= bytes;
1165                if (bytes <= 26)
1166                        line[0] = bytes + 'A' - 1;
1167                else
1168                        line[0] = bytes - 26 + 'a' - 1;
1169                encode_85(line + 1, cp, bytes);
1170                cp = (char *) cp + bytes;
1171                puts(line);
1172        }
1173        printf("\n");
1174        free(data);
1175}
1176
1177static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1178{
1179        printf("GIT binary patch\n");
1180        emit_binary_diff_body(one, two);
1181        emit_binary_diff_body(two, one);
1182}
1183
1184static void setup_diff_attr_check(struct git_attr_check *check)
1185{
1186        static struct git_attr *attr_diff;
1187
1188        if (!attr_diff) {
1189                attr_diff = git_attr("diff", 4);
1190        }
1191        check[0].attr = attr_diff;
1192}
1193
1194static void diff_filespec_check_attr(struct diff_filespec *one)
1195{
1196        struct git_attr_check attr_diff_check;
1197        int check_from_data = 0;
1198
1199        if (one->checked_attr)
1200                return;
1201
1202        setup_diff_attr_check(&attr_diff_check);
1203        one->is_binary = 0;
1204        one->funcname_pattern_ident = NULL;
1205
1206        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1207                const char *value;
1208
1209                /* binaryness */
1210                value = attr_diff_check.value;
1211                if (ATTR_TRUE(value))
1212                        ;
1213                else if (ATTR_FALSE(value))
1214                        one->is_binary = 1;
1215                else
1216                        check_from_data = 1;
1217
1218                /* funcname pattern ident */
1219                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1220                        ;
1221                else
1222                        one->funcname_pattern_ident = value;
1223        }
1224
1225        if (check_from_data) {
1226                if (!one->data && DIFF_FILE_VALID(one))
1227                        diff_populate_filespec(one, 0);
1228
1229                if (one->data)
1230                        one->is_binary = buffer_is_binary(one->data, one->size);
1231        }
1232}
1233
1234int diff_filespec_is_binary(struct diff_filespec *one)
1235{
1236        diff_filespec_check_attr(one);
1237        return one->is_binary;
1238}
1239
1240static const char *funcname_pattern(const char *ident)
1241{
1242        struct funcname_pattern *pp;
1243
1244        read_config_if_needed();
1245        for (pp = funcname_pattern_list; pp; pp = pp->next)
1246                if (!strcmp(ident, pp->name))
1247                        return pp->pattern;
1248        return NULL;
1249}
1250
1251static struct builtin_funcname_pattern {
1252        const char *name;
1253        const char *pattern;
1254} builtin_funcname_pattern[] = {
1255        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1256                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1257                        "^[     ]*\\(\\([       ]*"
1258                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1259                        "[      ]*([^;]*$\\)" },
1260        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1261};
1262
1263static const char *diff_funcname_pattern(struct diff_filespec *one)
1264{
1265        const char *ident, *pattern;
1266        int i;
1267
1268        diff_filespec_check_attr(one);
1269        ident = one->funcname_pattern_ident;
1270
1271        if (!ident)
1272                /*
1273                 * If the config file has "funcname.default" defined, that
1274                 * regexp is used; otherwise NULL is returned and xemit uses
1275                 * the built-in default.
1276                 */
1277                return funcname_pattern("default");
1278
1279        /* Look up custom "funcname.$ident" regexp from config. */
1280        pattern = funcname_pattern(ident);
1281        if (pattern)
1282                return pattern;
1283
1284        /*
1285         * And define built-in fallback patterns here.  Note that
1286         * these can be overriden by the user's config settings.
1287         */
1288        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1289                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1290                        return builtin_funcname_pattern[i].pattern;
1291
1292        return NULL;
1293}
1294
1295static void builtin_diff(const char *name_a,
1296                         const char *name_b,
1297                         struct diff_filespec *one,
1298                         struct diff_filespec *two,
1299                         const char *xfrm_msg,
1300                         struct diff_options *o,
1301                         int complete_rewrite)
1302{
1303        mmfile_t mf1, mf2;
1304        const char *lbl[2];
1305        char *a_one, *b_two;
1306        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1307        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1308
1309        a_one = quote_two("a/", name_a + (*name_a == '/'));
1310        b_two = quote_two("b/", name_b + (*name_b == '/'));
1311        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1312        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1313        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1314        if (lbl[0][0] == '/') {
1315                /* /dev/null */
1316                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1317                if (xfrm_msg && xfrm_msg[0])
1318                        printf("%s%s%s\n", set, xfrm_msg, reset);
1319        }
1320        else if (lbl[1][0] == '/') {
1321                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1322                if (xfrm_msg && xfrm_msg[0])
1323                        printf("%s%s%s\n", set, xfrm_msg, reset);
1324        }
1325        else {
1326                if (one->mode != two->mode) {
1327                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1328                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1329                }
1330                if (xfrm_msg && xfrm_msg[0])
1331                        printf("%s%s%s\n", set, xfrm_msg, reset);
1332                /*
1333                 * we do not run diff between different kind
1334                 * of objects.
1335                 */
1336                if ((one->mode ^ two->mode) & S_IFMT)
1337                        goto free_ab_and_return;
1338                if (complete_rewrite) {
1339                        emit_rewrite_diff(name_a, name_b, one, two,
1340                                        DIFF_OPT_TST(o, COLOR_DIFF));
1341                        o->found_changes = 1;
1342                        goto free_ab_and_return;
1343                }
1344        }
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_OPT_TST(o, TEXT) &&
1350            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1351                /* Quite common confusing case */
1352                if (mf1.size == mf2.size &&
1353                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1354                        goto free_ab_and_return;
1355                if (DIFF_OPT_TST(o, BINARY))
1356                        emit_binary_diff(&mf1, &mf2);
1357                else
1358                        printf("Binary files %s and %s differ\n",
1359                               lbl[0], lbl[1]);
1360                o->found_changes = 1;
1361        }
1362        else {
1363                /* Crazy xdl interfaces.. */
1364                const char *diffopts = getenv("GIT_DIFF_OPTS");
1365                xpparam_t xpp;
1366                xdemitconf_t xecfg;
1367                xdemitcb_t ecb;
1368                struct emit_callback ecbdata;
1369                const char *funcname_pattern;
1370
1371                funcname_pattern = diff_funcname_pattern(one);
1372                if (!funcname_pattern)
1373                        funcname_pattern = diff_funcname_pattern(two);
1374
1375                memset(&xecfg, 0, sizeof(xecfg));
1376                memset(&ecbdata, 0, sizeof(ecbdata));
1377                ecbdata.label_path = lbl;
1378                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1379                ecbdata.found_changesp = &o->found_changes;
1380                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1381                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1382                xecfg.ctxlen = o->context;
1383                xecfg.flags = XDL_EMIT_FUNCNAMES;
1384                if (funcname_pattern)
1385                        xdiff_set_find_func(&xecfg, funcname_pattern);
1386                if (!diffopts)
1387                        ;
1388                else if (!prefixcmp(diffopts, "--unified="))
1389                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1390                else if (!prefixcmp(diffopts, "-u"))
1391                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1392                ecb.outf = xdiff_outf;
1393                ecb.priv = &ecbdata;
1394                ecbdata.xm.consume = fn_out_consume;
1395                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1396                        ecbdata.diff_words =
1397                                xcalloc(1, sizeof(struct diff_words_data));
1398                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1399                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1400                        free_diff_words_data(&ecbdata);
1401        }
1402
1403 free_ab_and_return:
1404        diff_free_filespec_data(one);
1405        diff_free_filespec_data(two);
1406        free(a_one);
1407        free(b_two);
1408        return;
1409}
1410
1411static void builtin_diffstat(const char *name_a, const char *name_b,
1412                             struct diff_filespec *one,
1413                             struct diff_filespec *two,
1414                             struct diffstat_t *diffstat,
1415                             struct diff_options *o,
1416                             int complete_rewrite)
1417{
1418        mmfile_t mf1, mf2;
1419        struct diffstat_file *data;
1420
1421        data = diffstat_add(diffstat, name_a, name_b);
1422
1423        if (!one || !two) {
1424                data->is_unmerged = 1;
1425                return;
1426        }
1427        if (complete_rewrite) {
1428                diff_populate_filespec(one, 0);
1429                diff_populate_filespec(two, 0);
1430                data->deleted = count_lines(one->data, one->size);
1431                data->added = count_lines(two->data, two->size);
1432                goto free_and_return;
1433        }
1434        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1435                die("unable to read files to diff");
1436
1437        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1438                data->is_binary = 1;
1439                data->added = mf2.size;
1440                data->deleted = mf1.size;
1441        } else {
1442                /* Crazy xdl interfaces.. */
1443                xpparam_t xpp;
1444                xdemitconf_t xecfg;
1445                xdemitcb_t ecb;
1446
1447                memset(&xecfg, 0, sizeof(xecfg));
1448                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1449                ecb.outf = xdiff_outf;
1450                ecb.priv = diffstat;
1451                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1452        }
1453
1454 free_and_return:
1455        diff_free_filespec_data(one);
1456        diff_free_filespec_data(two);
1457}
1458
1459static void builtin_checkdiff(const char *name_a, const char *name_b,
1460                             struct diff_filespec *one,
1461                             struct diff_filespec *two, struct diff_options *o)
1462{
1463        mmfile_t mf1, mf2;
1464        struct checkdiff_t data;
1465
1466        if (!two)
1467                return;
1468
1469        memset(&data, 0, sizeof(data));
1470        data.xm.consume = checkdiff_consume;
1471        data.filename = name_b ? name_b : name_a;
1472        data.lineno = 0;
1473        data.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1474        data.ws_rule = whitespace_rule(data.filename);
1475
1476        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1477                die("unable to read files to diff");
1478
1479        if (diff_filespec_is_binary(two))
1480                goto free_and_return;
1481        else {
1482                /* Crazy xdl interfaces.. */
1483                xpparam_t xpp;
1484                xdemitconf_t xecfg;
1485                xdemitcb_t ecb;
1486
1487                memset(&xecfg, 0, sizeof(xecfg));
1488                xpp.flags = XDF_NEED_MINIMAL;
1489                ecb.outf = xdiff_outf;
1490                ecb.priv = &data;
1491                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1492        }
1493 free_and_return:
1494        diff_free_filespec_data(one);
1495        diff_free_filespec_data(two);
1496        if (data.status)
1497                DIFF_OPT_SET(o, CHECK_FAILED);
1498}
1499
1500struct diff_filespec *alloc_filespec(const char *path)
1501{
1502        int namelen = strlen(path);
1503        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1504
1505        memset(spec, 0, sizeof(*spec));
1506        spec->path = (char *)(spec + 1);
1507        memcpy(spec->path, path, namelen+1);
1508        spec->count = 1;
1509        return spec;
1510}
1511
1512void free_filespec(struct diff_filespec *spec)
1513{
1514        if (!--spec->count) {
1515                diff_free_filespec_data(spec);
1516                free(spec);
1517        }
1518}
1519
1520void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1521                   unsigned short mode)
1522{
1523        if (mode) {
1524                spec->mode = canon_mode(mode);
1525                hashcpy(spec->sha1, sha1);
1526                spec->sha1_valid = !is_null_sha1(sha1);
1527        }
1528}
1529
1530/*
1531 * Given a name and sha1 pair, if the index tells us the file in
1532 * the work tree has that object contents, return true, so that
1533 * prepare_temp_file() does not have to inflate and extract.
1534 */
1535static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1536{
1537        struct cache_entry *ce;
1538        struct stat st;
1539        int pos, len;
1540
1541        /* We do not read the cache ourselves here, because the
1542         * benchmark with my previous version that always reads cache
1543         * shows that it makes things worse for diff-tree comparing
1544         * two linux-2.6 kernel trees in an already checked out work
1545         * tree.  This is because most diff-tree comparisons deal with
1546         * only a small number of files, while reading the cache is
1547         * expensive for a large project, and its cost outweighs the
1548         * savings we get by not inflating the object to a temporary
1549         * file.  Practically, this code only helps when we are used
1550         * by diff-cache --cached, which does read the cache before
1551         * calling us.
1552         */
1553        if (!active_cache)
1554                return 0;
1555
1556        /* We want to avoid the working directory if our caller
1557         * doesn't need the data in a normal file, this system
1558         * is rather slow with its stat/open/mmap/close syscalls,
1559         * and the object is contained in a pack file.  The pack
1560         * is probably already open and will be faster to obtain
1561         * the data through than the working directory.  Loose
1562         * objects however would tend to be slower as they need
1563         * to be individually opened and inflated.
1564         */
1565        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1566                return 0;
1567
1568        len = strlen(name);
1569        pos = cache_name_pos(name, len);
1570        if (pos < 0)
1571                return 0;
1572        ce = active_cache[pos];
1573        if ((lstat(name, &st) < 0) ||
1574            !S_ISREG(st.st_mode) || /* careful! */
1575            ce_match_stat(ce, &st, 0) ||
1576            hashcmp(sha1, ce->sha1))
1577                return 0;
1578        /* we return 1 only when we can stat, it is a regular file,
1579         * stat information matches, and sha1 recorded in the cache
1580         * matches.  I.e. we know the file in the work tree really is
1581         * the same as the <name, sha1> pair.
1582         */
1583        return 1;
1584}
1585
1586static int populate_from_stdin(struct diff_filespec *s)
1587{
1588        struct strbuf buf;
1589        size_t size = 0;
1590
1591        strbuf_init(&buf, 0);
1592        if (strbuf_read(&buf, 0, 0) < 0)
1593                return error("error while reading from stdin %s",
1594                                     strerror(errno));
1595
1596        s->should_munmap = 0;
1597        s->data = strbuf_detach(&buf, &size);
1598        s->size = size;
1599        s->should_free = 1;
1600        return 0;
1601}
1602
1603static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1604{
1605        int len;
1606        char *data = xmalloc(100);
1607        len = snprintf(data, 100,
1608                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1609        s->data = data;
1610        s->size = len;
1611        s->should_free = 1;
1612        if (size_only) {
1613                s->data = NULL;
1614                free(data);
1615        }
1616        return 0;
1617}
1618
1619/*
1620 * While doing rename detection and pickaxe operation, we may need to
1621 * grab the data for the blob (or file) for our own in-core comparison.
1622 * diff_filespec has data and size fields for this purpose.
1623 */
1624int diff_populate_filespec(struct diff_filespec *s, int size_only)
1625{
1626        int err = 0;
1627        if (!DIFF_FILE_VALID(s))
1628                die("internal error: asking to populate invalid file.");
1629        if (S_ISDIR(s->mode))
1630                return -1;
1631
1632        if (s->data)
1633                return 0;
1634
1635        if (size_only && 0 < s->size)
1636                return 0;
1637
1638        if (S_ISGITLINK(s->mode))
1639                return diff_populate_gitlink(s, size_only);
1640
1641        if (!s->sha1_valid ||
1642            reuse_worktree_file(s->path, s->sha1, 0)) {
1643                struct strbuf buf;
1644                struct stat st;
1645                int fd;
1646
1647                if (!strcmp(s->path, "-"))
1648                        return populate_from_stdin(s);
1649
1650                if (lstat(s->path, &st) < 0) {
1651                        if (errno == ENOENT) {
1652                        err_empty:
1653                                err = -1;
1654                        empty:
1655                                s->data = (char *)"";
1656                                s->size = 0;
1657                                return err;
1658                        }
1659                }
1660                s->size = xsize_t(st.st_size);
1661                if (!s->size)
1662                        goto empty;
1663                if (size_only)
1664                        return 0;
1665                if (S_ISLNK(st.st_mode)) {
1666                        int ret;
1667                        s->data = xmalloc(s->size);
1668                        s->should_free = 1;
1669                        ret = readlink(s->path, s->data, s->size);
1670                        if (ret < 0) {
1671                                free(s->data);
1672                                goto err_empty;
1673                        }
1674                        return 0;
1675                }
1676                fd = open(s->path, O_RDONLY);
1677                if (fd < 0)
1678                        goto err_empty;
1679                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1680                close(fd);
1681                s->should_munmap = 1;
1682
1683                /*
1684                 * Convert from working tree format to canonical git format
1685                 */
1686                strbuf_init(&buf, 0);
1687                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1688                        size_t size = 0;
1689                        munmap(s->data, s->size);
1690                        s->should_munmap = 0;
1691                        s->data = strbuf_detach(&buf, &size);
1692                        s->size = size;
1693                        s->should_free = 1;
1694                }
1695        }
1696        else {
1697                enum object_type type;
1698                if (size_only)
1699                        type = sha1_object_info(s->sha1, &s->size);
1700                else {
1701                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1702                        s->should_free = 1;
1703                }
1704        }
1705        return 0;
1706}
1707
1708void diff_free_filespec_blob(struct diff_filespec *s)
1709{
1710        if (s->should_free)
1711                free(s->data);
1712        else if (s->should_munmap)
1713                munmap(s->data, s->size);
1714
1715        if (s->should_free || s->should_munmap) {
1716                s->should_free = s->should_munmap = 0;
1717                s->data = NULL;
1718        }
1719}
1720
1721void diff_free_filespec_data(struct diff_filespec *s)
1722{
1723        diff_free_filespec_blob(s);
1724        free(s->cnt_data);
1725        s->cnt_data = NULL;
1726}
1727
1728static void prep_temp_blob(struct diff_tempfile *temp,
1729                           void *blob,
1730                           unsigned long size,
1731                           const unsigned char *sha1,
1732                           int mode)
1733{
1734        int fd;
1735
1736        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1737        if (fd < 0)
1738                die("unable to create temp-file: %s", strerror(errno));
1739        if (write_in_full(fd, blob, size) != size)
1740                die("unable to write temp-file");
1741        close(fd);
1742        temp->name = temp->tmp_path;
1743        strcpy(temp->hex, sha1_to_hex(sha1));
1744        temp->hex[40] = 0;
1745        sprintf(temp->mode, "%06o", mode);
1746}
1747
1748static void prepare_temp_file(const char *name,
1749                              struct diff_tempfile *temp,
1750                              struct diff_filespec *one)
1751{
1752        if (!DIFF_FILE_VALID(one)) {
1753        not_a_valid_file:
1754                /* A '-' entry produces this for file-2, and
1755                 * a '+' entry produces this for file-1.
1756                 */
1757                temp->name = "/dev/null";
1758                strcpy(temp->hex, ".");
1759                strcpy(temp->mode, ".");
1760                return;
1761        }
1762
1763        if (!one->sha1_valid ||
1764            reuse_worktree_file(name, one->sha1, 1)) {
1765                struct stat st;
1766                if (lstat(name, &st) < 0) {
1767                        if (errno == ENOENT)
1768                                goto not_a_valid_file;
1769                        die("stat(%s): %s", name, strerror(errno));
1770                }
1771                if (S_ISLNK(st.st_mode)) {
1772                        int ret;
1773                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1774                        size_t sz = xsize_t(st.st_size);
1775                        if (sizeof(buf) <= st.st_size)
1776                                die("symlink too long: %s", name);
1777                        ret = readlink(name, buf, sz);
1778                        if (ret < 0)
1779                                die("readlink(%s)", name);
1780                        prep_temp_blob(temp, buf, sz,
1781                                       (one->sha1_valid ?
1782                                        one->sha1 : null_sha1),
1783                                       (one->sha1_valid ?
1784                                        one->mode : S_IFLNK));
1785                }
1786                else {
1787                        /* we can borrow from the file in the work tree */
1788                        temp->name = name;
1789                        if (!one->sha1_valid)
1790                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1791                        else
1792                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1793                        /* Even though we may sometimes borrow the
1794                         * contents from the work tree, we always want
1795                         * one->mode.  mode is trustworthy even when
1796                         * !(one->sha1_valid), as long as
1797                         * DIFF_FILE_VALID(one).
1798                         */
1799                        sprintf(temp->mode, "%06o", one->mode);
1800                }
1801                return;
1802        }
1803        else {
1804                if (diff_populate_filespec(one, 0))
1805                        die("cannot read data blob for %s", one->path);
1806                prep_temp_blob(temp, one->data, one->size,
1807                               one->sha1, one->mode);
1808        }
1809}
1810
1811static void remove_tempfile(void)
1812{
1813        int i;
1814
1815        for (i = 0; i < 2; i++)
1816                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1817                        unlink(diff_temp[i].name);
1818                        diff_temp[i].name = NULL;
1819                }
1820}
1821
1822static void remove_tempfile_on_signal(int signo)
1823{
1824        remove_tempfile();
1825        signal(SIGINT, SIG_DFL);
1826        raise(signo);
1827}
1828
1829/* An external diff command takes:
1830 *
1831 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1832 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1833 *
1834 */
1835static void run_external_diff(const char *pgm,
1836                              const char *name,
1837                              const char *other,
1838                              struct diff_filespec *one,
1839                              struct diff_filespec *two,
1840                              const char *xfrm_msg,
1841                              int complete_rewrite)
1842{
1843        const char *spawn_arg[10];
1844        struct diff_tempfile *temp = diff_temp;
1845        int retval;
1846        static int atexit_asked = 0;
1847        const char *othername;
1848        const char **arg = &spawn_arg[0];
1849
1850        othername = (other? other : name);
1851        if (one && two) {
1852                prepare_temp_file(name, &temp[0], one);
1853                prepare_temp_file(othername, &temp[1], two);
1854                if (! atexit_asked &&
1855                    (temp[0].name == temp[0].tmp_path ||
1856                     temp[1].name == temp[1].tmp_path)) {
1857                        atexit_asked = 1;
1858                        atexit(remove_tempfile);
1859                }
1860                signal(SIGINT, remove_tempfile_on_signal);
1861        }
1862
1863        if (one && two) {
1864                *arg++ = pgm;
1865                *arg++ = name;
1866                *arg++ = temp[0].name;
1867                *arg++ = temp[0].hex;
1868                *arg++ = temp[0].mode;
1869                *arg++ = temp[1].name;
1870                *arg++ = temp[1].hex;
1871                *arg++ = temp[1].mode;
1872                if (other) {
1873                        *arg++ = other;
1874                        *arg++ = xfrm_msg;
1875                }
1876        } else {
1877                *arg++ = pgm;
1878                *arg++ = name;
1879        }
1880        *arg = NULL;
1881        fflush(NULL);
1882        retval = run_command_v_opt(spawn_arg, 0);
1883        remove_tempfile();
1884        if (retval) {
1885                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1886                exit(1);
1887        }
1888}
1889
1890static const char *external_diff_attr(const char *name)
1891{
1892        struct git_attr_check attr_diff_check;
1893
1894        setup_diff_attr_check(&attr_diff_check);
1895        if (!git_checkattr(name, 1, &attr_diff_check)) {
1896                const char *value = attr_diff_check.value;
1897                if (!ATTR_TRUE(value) &&
1898                    !ATTR_FALSE(value) &&
1899                    !ATTR_UNSET(value)) {
1900                        struct ll_diff_driver *drv;
1901
1902                        read_config_if_needed();
1903                        for (drv = user_diff; drv; drv = drv->next)
1904                                if (!strcmp(drv->name, value))
1905                                        return drv->cmd;
1906                }
1907        }
1908        return NULL;
1909}
1910
1911static void run_diff_cmd(const char *pgm,
1912                         const char *name,
1913                         const char *other,
1914                         struct diff_filespec *one,
1915                         struct diff_filespec *two,
1916                         const char *xfrm_msg,
1917                         struct diff_options *o,
1918                         int complete_rewrite)
1919{
1920        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
1921                pgm = NULL;
1922        else {
1923                const char *cmd = external_diff_attr(name);
1924                if (cmd)
1925                        pgm = cmd;
1926        }
1927
1928        if (pgm) {
1929                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1930                                  complete_rewrite);
1931                return;
1932        }
1933        if (one && two)
1934                builtin_diff(name, other ? other : name,
1935                             one, two, xfrm_msg, o, complete_rewrite);
1936        else
1937                printf("* Unmerged path %s\n", name);
1938}
1939
1940static void diff_fill_sha1_info(struct diff_filespec *one)
1941{
1942        if (DIFF_FILE_VALID(one)) {
1943                if (!one->sha1_valid) {
1944                        struct stat st;
1945                        if (!strcmp(one->path, "-")) {
1946                                hashcpy(one->sha1, null_sha1);
1947                                return;
1948                        }
1949                        if (lstat(one->path, &st) < 0)
1950                                die("stat %s", one->path);
1951                        if (index_path(one->sha1, one->path, &st, 0))
1952                                die("cannot hash %s\n", one->path);
1953                }
1954        }
1955        else
1956                hashclr(one->sha1);
1957}
1958
1959static int similarity_index(struct diff_filepair *p)
1960{
1961        return p->score * 100 / MAX_SCORE;
1962}
1963
1964static void run_diff(struct diff_filepair *p, struct diff_options *o)
1965{
1966        const char *pgm = external_diff();
1967        struct strbuf msg;
1968        char *xfrm_msg;
1969        struct diff_filespec *one = p->one;
1970        struct diff_filespec *two = p->two;
1971        const char *name;
1972        const char *other;
1973        int complete_rewrite = 0;
1974
1975
1976        if (DIFF_PAIR_UNMERGED(p)) {
1977                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1978                return;
1979        }
1980
1981        name  = p->one->path;
1982        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1983        diff_fill_sha1_info(one);
1984        diff_fill_sha1_info(two);
1985
1986        strbuf_init(&msg, PATH_MAX * 2 + 300);
1987        switch (p->status) {
1988        case DIFF_STATUS_COPIED:
1989                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1990                strbuf_addstr(&msg, "\ncopy from ");
1991                quote_c_style(name, &msg, NULL, 0);
1992                strbuf_addstr(&msg, "\ncopy to ");
1993                quote_c_style(other, &msg, NULL, 0);
1994                strbuf_addch(&msg, '\n');
1995                break;
1996        case DIFF_STATUS_RENAMED:
1997                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1998                strbuf_addstr(&msg, "\nrename from ");
1999                quote_c_style(name, &msg, NULL, 0);
2000                strbuf_addstr(&msg, "\nrename to ");
2001                quote_c_style(other, &msg, NULL, 0);
2002                strbuf_addch(&msg, '\n');
2003                break;
2004        case DIFF_STATUS_MODIFIED:
2005                if (p->score) {
2006                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
2007                                        similarity_index(p));
2008                        complete_rewrite = 1;
2009                        break;
2010                }
2011                /* fallthru */
2012        default:
2013                /* nothing */
2014                ;
2015        }
2016
2017        if (hashcmp(one->sha1, two->sha1)) {
2018                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2019
2020                if (DIFF_OPT_TST(o, BINARY)) {
2021                        mmfile_t mf;
2022                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2023                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2024                                abbrev = 40;
2025                }
2026                strbuf_addf(&msg, "index %.*s..%.*s",
2027                                abbrev, sha1_to_hex(one->sha1),
2028                                abbrev, sha1_to_hex(two->sha1));
2029                if (one->mode == two->mode)
2030                        strbuf_addf(&msg, " %06o", one->mode);
2031                strbuf_addch(&msg, '\n');
2032        }
2033
2034        if (msg.len)
2035                strbuf_setlen(&msg, msg.len - 1);
2036        xfrm_msg = msg.len ? msg.buf : NULL;
2037
2038        if (!pgm &&
2039            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2040            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2041                /* a filepair that changes between file and symlink
2042                 * needs to be split into deletion and creation.
2043                 */
2044                struct diff_filespec *null = alloc_filespec(two->path);
2045                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
2046                free(null);
2047                null = alloc_filespec(one->path);
2048                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
2049                free(null);
2050        }
2051        else
2052                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
2053                             complete_rewrite);
2054
2055        strbuf_release(&msg);
2056}
2057
2058static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2059                         struct diffstat_t *diffstat)
2060{
2061        const char *name;
2062        const char *other;
2063        int complete_rewrite = 0;
2064
2065        if (DIFF_PAIR_UNMERGED(p)) {
2066                /* unmerged */
2067                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2068                return;
2069        }
2070
2071        name = p->one->path;
2072        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2073
2074        diff_fill_sha1_info(p->one);
2075        diff_fill_sha1_info(p->two);
2076
2077        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2078                complete_rewrite = 1;
2079        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2080}
2081
2082static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2083{
2084        const char *name;
2085        const char *other;
2086
2087        if (DIFF_PAIR_UNMERGED(p)) {
2088                /* unmerged */
2089                return;
2090        }
2091
2092        name = p->one->path;
2093        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2094
2095        diff_fill_sha1_info(p->one);
2096        diff_fill_sha1_info(p->two);
2097
2098        builtin_checkdiff(name, other, p->one, p->two, o);
2099}
2100
2101void diff_setup(struct diff_options *options)
2102{
2103        memset(options, 0, sizeof(*options));
2104        options->line_termination = '\n';
2105        options->break_opt = -1;
2106        options->rename_limit = -1;
2107        options->context = 3;
2108        options->msg_sep = "";
2109
2110        options->change = diff_change;
2111        options->add_remove = diff_addremove;
2112        if (diff_use_color_default)
2113                DIFF_OPT_SET(options, COLOR_DIFF);
2114        else
2115                DIFF_OPT_CLR(options, COLOR_DIFF);
2116        options->detect_rename = diff_detect_rename_default;
2117}
2118
2119int diff_setup_done(struct diff_options *options)
2120{
2121        int count = 0;
2122
2123        if (options->output_format & DIFF_FORMAT_NAME)
2124                count++;
2125        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2126                count++;
2127        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2128                count++;
2129        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2130                count++;
2131        if (count > 1)
2132                die("--name-only, --name-status, --check and -s are mutually exclusive");
2133
2134        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2135                options->detect_rename = DIFF_DETECT_COPY;
2136
2137        if (options->output_format & (DIFF_FORMAT_NAME |
2138                                      DIFF_FORMAT_NAME_STATUS |
2139                                      DIFF_FORMAT_CHECKDIFF |
2140                                      DIFF_FORMAT_NO_OUTPUT))
2141                options->output_format &= ~(DIFF_FORMAT_RAW |
2142                                            DIFF_FORMAT_NUMSTAT |
2143                                            DIFF_FORMAT_DIFFSTAT |
2144                                            DIFF_FORMAT_SHORTSTAT |
2145                                            DIFF_FORMAT_SUMMARY |
2146                                            DIFF_FORMAT_PATCH);
2147
2148        /*
2149         * These cases always need recursive; we do not drop caller-supplied
2150         * recursive bits for other formats here.
2151         */
2152        if (options->output_format & (DIFF_FORMAT_PATCH |
2153                                      DIFF_FORMAT_NUMSTAT |
2154                                      DIFF_FORMAT_DIFFSTAT |
2155                                      DIFF_FORMAT_SHORTSTAT |
2156                                      DIFF_FORMAT_SUMMARY |
2157                                      DIFF_FORMAT_CHECKDIFF))
2158                DIFF_OPT_SET(options, RECURSIVE);
2159        /*
2160         * Also pickaxe would not work very well if you do not say recursive
2161         */
2162        if (options->pickaxe)
2163                DIFF_OPT_SET(options, RECURSIVE);
2164
2165        if (options->detect_rename && options->rename_limit < 0)
2166                options->rename_limit = diff_rename_limit_default;
2167        if (options->setup & DIFF_SETUP_USE_CACHE) {
2168                if (!active_cache)
2169                        /* read-cache does not die even when it fails
2170                         * so it is safe for us to do this here.  Also
2171                         * it does not smudge active_cache or active_nr
2172                         * when it fails, so we do not have to worry about
2173                         * cleaning it up ourselves either.
2174                         */
2175                        read_cache();
2176        }
2177        if (options->abbrev <= 0 || 40 < options->abbrev)
2178                options->abbrev = 40; /* full */
2179
2180        /*
2181         * It does not make sense to show the first hit we happened
2182         * to have found.  It does not make sense not to return with
2183         * exit code in such a case either.
2184         */
2185        if (DIFF_OPT_TST(options, QUIET)) {
2186                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2187                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2188        }
2189
2190        /*
2191         * If we postprocess in diffcore, we cannot simply return
2192         * upon the first hit.  We need to run diff as usual.
2193         */
2194        if (options->pickaxe || options->filter)
2195                DIFF_OPT_CLR(options, QUIET);
2196
2197        return 0;
2198}
2199
2200static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2201{
2202        char c, *eq;
2203        int len;
2204
2205        if (*arg != '-')
2206                return 0;
2207        c = *++arg;
2208        if (!c)
2209                return 0;
2210        if (c == arg_short) {
2211                c = *++arg;
2212                if (!c)
2213                        return 1;
2214                if (val && isdigit(c)) {
2215                        char *end;
2216                        int n = strtoul(arg, &end, 10);
2217                        if (*end)
2218                                return 0;
2219                        *val = n;
2220                        return 1;
2221                }
2222                return 0;
2223        }
2224        if (c != '-')
2225                return 0;
2226        arg++;
2227        eq = strchr(arg, '=');
2228        if (eq)
2229                len = eq - arg;
2230        else
2231                len = strlen(arg);
2232        if (!len || strncmp(arg, arg_long, len))
2233                return 0;
2234        if (eq) {
2235                int n;
2236                char *end;
2237                if (!isdigit(*++eq))
2238                        return 0;
2239                n = strtoul(eq, &end, 10);
2240                if (*end)
2241                        return 0;
2242                *val = n;
2243        }
2244        return 1;
2245}
2246
2247static int diff_scoreopt_parse(const char *opt);
2248
2249int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2250{
2251        const char *arg = av[0];
2252
2253        /* Output format options */
2254        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2255                options->output_format |= DIFF_FORMAT_PATCH;
2256        else if (opt_arg(arg, 'U', "unified", &options->context))
2257                options->output_format |= DIFF_FORMAT_PATCH;
2258        else if (!strcmp(arg, "--raw"))
2259                options->output_format |= DIFF_FORMAT_RAW;
2260        else if (!strcmp(arg, "--patch-with-raw"))
2261                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2262        else if (!strcmp(arg, "--numstat"))
2263                options->output_format |= DIFF_FORMAT_NUMSTAT;
2264        else if (!strcmp(arg, "--shortstat"))
2265                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2266        else if (!strcmp(arg, "--check"))
2267                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2268        else if (!strcmp(arg, "--summary"))
2269                options->output_format |= DIFF_FORMAT_SUMMARY;
2270        else if (!strcmp(arg, "--patch-with-stat"))
2271                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2272        else if (!strcmp(arg, "--name-only"))
2273                options->output_format |= DIFF_FORMAT_NAME;
2274        else if (!strcmp(arg, "--name-status"))
2275                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2276        else if (!strcmp(arg, "-s"))
2277                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2278        else if (!prefixcmp(arg, "--stat")) {
2279                char *end;
2280                int width = options->stat_width;
2281                int name_width = options->stat_name_width;
2282                arg += 6;
2283                end = (char *)arg;
2284
2285                switch (*arg) {
2286                case '-':
2287                        if (!prefixcmp(arg, "-width="))
2288                                width = strtoul(arg + 7, &end, 10);
2289                        else if (!prefixcmp(arg, "-name-width="))
2290                                name_width = strtoul(arg + 12, &end, 10);
2291                        break;
2292                case '=':
2293                        width = strtoul(arg+1, &end, 10);
2294                        if (*end == ',')
2295                                name_width = strtoul(end+1, &end, 10);
2296                }
2297
2298                /* Important! This checks all the error cases! */
2299                if (*end)
2300                        return 0;
2301                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2302                options->stat_name_width = name_width;
2303                options->stat_width = width;
2304        }
2305
2306        /* renames options */
2307        else if (!prefixcmp(arg, "-B")) {
2308                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2309                        return -1;
2310        }
2311        else if (!prefixcmp(arg, "-M")) {
2312                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2313                        return -1;
2314                options->detect_rename = DIFF_DETECT_RENAME;
2315        }
2316        else if (!prefixcmp(arg, "-C")) {
2317                if (options->detect_rename == DIFF_DETECT_COPY)
2318                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2319                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2320                        return -1;
2321                options->detect_rename = DIFF_DETECT_COPY;
2322        }
2323        else if (!strcmp(arg, "--no-renames"))
2324                options->detect_rename = 0;
2325
2326        /* xdiff options */
2327        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2328                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2329        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2330                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2331        else if (!strcmp(arg, "--ignore-space-at-eol"))
2332                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2333
2334        /* flags options */
2335        else if (!strcmp(arg, "--binary")) {
2336                options->output_format |= DIFF_FORMAT_PATCH;
2337                DIFF_OPT_SET(options, BINARY);
2338        }
2339        else if (!strcmp(arg, "--full-index"))
2340                DIFF_OPT_SET(options, FULL_INDEX);
2341        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2342                DIFF_OPT_SET(options, TEXT);
2343        else if (!strcmp(arg, "-R"))
2344                DIFF_OPT_SET(options, REVERSE_DIFF);
2345        else if (!strcmp(arg, "--find-copies-harder"))
2346                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2347        else if (!strcmp(arg, "--follow"))
2348                DIFF_OPT_SET(options, FOLLOW_RENAMES);
2349        else if (!strcmp(arg, "--color"))
2350                DIFF_OPT_SET(options, COLOR_DIFF);
2351        else if (!strcmp(arg, "--no-color"))
2352                DIFF_OPT_CLR(options, COLOR_DIFF);
2353        else if (!strcmp(arg, "--color-words"))
2354                options->flags |= DIFF_OPT_COLOR_DIFF | DIFF_OPT_COLOR_DIFF_WORDS;
2355        else if (!strcmp(arg, "--exit-code"))
2356                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2357        else if (!strcmp(arg, "--quiet"))
2358                DIFF_OPT_SET(options, QUIET);
2359        else if (!strcmp(arg, "--ext-diff"))
2360                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2361        else if (!strcmp(arg, "--no-ext-diff"))
2362                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2363
2364        /* misc options */
2365        else if (!strcmp(arg, "-z"))
2366                options->line_termination = 0;
2367        else if (!prefixcmp(arg, "-l"))
2368                options->rename_limit = strtoul(arg+2, NULL, 10);
2369        else if (!prefixcmp(arg, "-S"))
2370                options->pickaxe = arg + 2;
2371        else if (!strcmp(arg, "--pickaxe-all"))
2372                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2373        else if (!strcmp(arg, "--pickaxe-regex"))
2374                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2375        else if (!prefixcmp(arg, "-O"))
2376                options->orderfile = arg + 2;
2377        else if (!prefixcmp(arg, "--diff-filter="))
2378                options->filter = arg + 14;
2379        else if (!strcmp(arg, "--abbrev"))
2380                options->abbrev = DEFAULT_ABBREV;
2381        else if (!prefixcmp(arg, "--abbrev=")) {
2382                options->abbrev = strtoul(arg + 9, NULL, 10);
2383                if (options->abbrev < MINIMUM_ABBREV)
2384                        options->abbrev = MINIMUM_ABBREV;
2385                else if (40 < options->abbrev)
2386                        options->abbrev = 40;
2387        }
2388        else
2389                return 0;
2390        return 1;
2391}
2392
2393static int parse_num(const char **cp_p)
2394{
2395        unsigned long num, scale;
2396        int ch, dot;
2397        const char *cp = *cp_p;
2398
2399        num = 0;
2400        scale = 1;
2401        dot = 0;
2402        for(;;) {
2403                ch = *cp;
2404                if ( !dot && ch == '.' ) {
2405                        scale = 1;
2406                        dot = 1;
2407                } else if ( ch == '%' ) {
2408                        scale = dot ? scale*100 : 100;
2409                        cp++;   /* % is always at the end */
2410                        break;
2411                } else if ( ch >= '0' && ch <= '9' ) {
2412                        if ( scale < 100000 ) {
2413                                scale *= 10;
2414                                num = (num*10) + (ch-'0');
2415                        }
2416                } else {
2417                        break;
2418                }
2419                cp++;
2420        }
2421        *cp_p = cp;
2422
2423        /* user says num divided by scale and we say internally that
2424         * is MAX_SCORE * num / scale.
2425         */
2426        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2427}
2428
2429static int diff_scoreopt_parse(const char *opt)
2430{
2431        int opt1, opt2, cmd;
2432
2433        if (*opt++ != '-')
2434                return -1;
2435        cmd = *opt++;
2436        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2437                return -1; /* that is not a -M, -C nor -B option */
2438
2439        opt1 = parse_num(&opt);
2440        if (cmd != 'B')
2441                opt2 = 0;
2442        else {
2443                if (*opt == 0)
2444                        opt2 = 0;
2445                else if (*opt != '/')
2446                        return -1; /* we expect -B80/99 or -B80 */
2447                else {
2448                        opt++;
2449                        opt2 = parse_num(&opt);
2450                }
2451        }
2452        if (*opt != 0)
2453                return -1;
2454        return opt1 | (opt2 << 16);
2455}
2456
2457struct diff_queue_struct diff_queued_diff;
2458
2459void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2460{
2461        if (queue->alloc <= queue->nr) {
2462                queue->alloc = alloc_nr(queue->alloc);
2463                queue->queue = xrealloc(queue->queue,
2464                                        sizeof(dp) * queue->alloc);
2465        }
2466        queue->queue[queue->nr++] = dp;
2467}
2468
2469struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2470                                 struct diff_filespec *one,
2471                                 struct diff_filespec *two)
2472{
2473        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2474        dp->one = one;
2475        dp->two = two;
2476        if (queue)
2477                diff_q(queue, dp);
2478        return dp;
2479}
2480
2481void diff_free_filepair(struct diff_filepair *p)
2482{
2483        free_filespec(p->one);
2484        free_filespec(p->two);
2485        free(p);
2486}
2487
2488/* This is different from find_unique_abbrev() in that
2489 * it stuffs the result with dots for alignment.
2490 */
2491const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2492{
2493        int abblen;
2494        const char *abbrev;
2495        if (len == 40)
2496                return sha1_to_hex(sha1);
2497
2498        abbrev = find_unique_abbrev(sha1, len);
2499        if (!abbrev)
2500                return sha1_to_hex(sha1);
2501        abblen = strlen(abbrev);
2502        if (abblen < 37) {
2503                static char hex[41];
2504                if (len < abblen && abblen <= len + 2)
2505                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2506                else
2507                        sprintf(hex, "%s...", abbrev);
2508                return hex;
2509        }
2510        return sha1_to_hex(sha1);
2511}
2512
2513static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2514{
2515        int line_termination = opt->line_termination;
2516        int inter_name_termination = line_termination ? '\t' : '\0';
2517
2518        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2519                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2520                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2521                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2522        }
2523        if (p->score) {
2524                printf("%c%03d%c", p->status, similarity_index(p),
2525                           inter_name_termination);
2526        } else {
2527                printf("%c%c", p->status, inter_name_termination);
2528        }
2529
2530        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2531                write_name_quoted(p->one->path, stdout, inter_name_termination);
2532                write_name_quoted(p->two->path, stdout, line_termination);
2533        } else {
2534                const char *path = p->one->mode ? p->one->path : p->two->path;
2535                write_name_quoted(path, stdout, line_termination);
2536        }
2537}
2538
2539int diff_unmodified_pair(struct diff_filepair *p)
2540{
2541        /* This function is written stricter than necessary to support
2542         * the currently implemented transformers, but the idea is to
2543         * let transformers to produce diff_filepairs any way they want,
2544         * and filter and clean them up here before producing the output.
2545         */
2546        struct diff_filespec *one = p->one, *two = p->two;
2547
2548        if (DIFF_PAIR_UNMERGED(p))
2549                return 0; /* unmerged is interesting */
2550
2551        /* deletion, addition, mode or type change
2552         * and rename are all interesting.
2553         */
2554        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2555            DIFF_PAIR_MODE_CHANGED(p) ||
2556            strcmp(one->path, two->path))
2557                return 0;
2558
2559        /* both are valid and point at the same path.  that is, we are
2560         * dealing with a change.
2561         */
2562        if (one->sha1_valid && two->sha1_valid &&
2563            !hashcmp(one->sha1, two->sha1))
2564                return 1; /* no change */
2565        if (!one->sha1_valid && !two->sha1_valid)
2566                return 1; /* both look at the same file on the filesystem. */
2567        return 0;
2568}
2569
2570static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2571{
2572        if (diff_unmodified_pair(p))
2573                return;
2574
2575        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2576            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2577                return; /* no tree diffs in patch format */
2578
2579        run_diff(p, o);
2580}
2581
2582static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2583                            struct diffstat_t *diffstat)
2584{
2585        if (diff_unmodified_pair(p))
2586                return;
2587
2588        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2589            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2590                return; /* no tree diffs in patch format */
2591
2592        run_diffstat(p, o, diffstat);
2593}
2594
2595static void diff_flush_checkdiff(struct diff_filepair *p,
2596                struct diff_options *o)
2597{
2598        if (diff_unmodified_pair(p))
2599                return;
2600
2601        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2602            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2603                return; /* no tree diffs in patch format */
2604
2605        run_checkdiff(p, o);
2606}
2607
2608int diff_queue_is_empty(void)
2609{
2610        struct diff_queue_struct *q = &diff_queued_diff;
2611        int i;
2612        for (i = 0; i < q->nr; i++)
2613                if (!diff_unmodified_pair(q->queue[i]))
2614                        return 0;
2615        return 1;
2616}
2617
2618#if DIFF_DEBUG
2619void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2620{
2621        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2622                x, one ? one : "",
2623                s->path,
2624                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2625                s->mode,
2626                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2627        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2628                x, one ? one : "",
2629                s->size, s->xfrm_flags);
2630}
2631
2632void diff_debug_filepair(const struct diff_filepair *p, int i)
2633{
2634        diff_debug_filespec(p->one, i, "one");
2635        diff_debug_filespec(p->two, i, "two");
2636        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2637                p->score, p->status ? p->status : '?',
2638                p->one->rename_used, p->broken_pair);
2639}
2640
2641void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2642{
2643        int i;
2644        if (msg)
2645                fprintf(stderr, "%s\n", msg);
2646        fprintf(stderr, "q->nr = %d\n", q->nr);
2647        for (i = 0; i < q->nr; i++) {
2648                struct diff_filepair *p = q->queue[i];
2649                diff_debug_filepair(p, i);
2650        }
2651}
2652#endif
2653
2654static void diff_resolve_rename_copy(void)
2655{
2656        int i;
2657        struct diff_filepair *p;
2658        struct diff_queue_struct *q = &diff_queued_diff;
2659
2660        diff_debug_queue("resolve-rename-copy", q);
2661
2662        for (i = 0; i < q->nr; i++) {
2663                p = q->queue[i];
2664                p->status = 0; /* undecided */
2665                if (DIFF_PAIR_UNMERGED(p))
2666                        p->status = DIFF_STATUS_UNMERGED;
2667                else if (!DIFF_FILE_VALID(p->one))
2668                        p->status = DIFF_STATUS_ADDED;
2669                else if (!DIFF_FILE_VALID(p->two))
2670                        p->status = DIFF_STATUS_DELETED;
2671                else if (DIFF_PAIR_TYPE_CHANGED(p))
2672                        p->status = DIFF_STATUS_TYPE_CHANGED;
2673
2674                /* from this point on, we are dealing with a pair
2675                 * whose both sides are valid and of the same type, i.e.
2676                 * either in-place edit or rename/copy edit.
2677                 */
2678                else if (DIFF_PAIR_RENAME(p)) {
2679                        /*
2680                         * A rename might have re-connected a broken
2681                         * pair up, causing the pathnames to be the
2682                         * same again. If so, that's not a rename at
2683                         * all, just a modification..
2684                         *
2685                         * Otherwise, see if this source was used for
2686                         * multiple renames, in which case we decrement
2687                         * the count, and call it a copy.
2688                         */
2689                        if (!strcmp(p->one->path, p->two->path))
2690                                p->status = DIFF_STATUS_MODIFIED;
2691                        else if (--p->one->rename_used > 0)
2692                                p->status = DIFF_STATUS_COPIED;
2693                        else
2694                                p->status = DIFF_STATUS_RENAMED;
2695                }
2696                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2697                         p->one->mode != p->two->mode ||
2698                         is_null_sha1(p->one->sha1))
2699                        p->status = DIFF_STATUS_MODIFIED;
2700                else {
2701                        /* This is a "no-change" entry and should not
2702                         * happen anymore, but prepare for broken callers.
2703                         */
2704                        error("feeding unmodified %s to diffcore",
2705                              p->one->path);
2706                        p->status = DIFF_STATUS_UNKNOWN;
2707                }
2708        }
2709        diff_debug_queue("resolve-rename-copy done", q);
2710}
2711
2712static int check_pair_status(struct diff_filepair *p)
2713{
2714        switch (p->status) {
2715        case DIFF_STATUS_UNKNOWN:
2716                return 0;
2717        case 0:
2718                die("internal error in diff-resolve-rename-copy");
2719        default:
2720                return 1;
2721        }
2722}
2723
2724static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2725{
2726        int fmt = opt->output_format;
2727
2728        if (fmt & DIFF_FORMAT_CHECKDIFF)
2729                diff_flush_checkdiff(p, opt);
2730        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2731                diff_flush_raw(p, opt);
2732        else if (fmt & DIFF_FORMAT_NAME)
2733                write_name_quoted(p->two->path, stdout, opt->line_termination);
2734}
2735
2736static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2737{
2738        if (fs->mode)
2739                printf(" %s mode %06o ", newdelete, fs->mode);
2740        else
2741                printf(" %s ", newdelete);
2742        write_name_quoted(fs->path, stdout, '\n');
2743}
2744
2745
2746static void show_mode_change(struct diff_filepair *p, int show_name)
2747{
2748        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2749                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2750                        show_name ? ' ' : '\n');
2751                if (show_name) {
2752                        write_name_quoted(p->two->path, stdout, '\n');
2753                }
2754        }
2755}
2756
2757static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2758{
2759        char *names = pprint_rename(p->one->path, p->two->path);
2760
2761        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2762        free(names);
2763        show_mode_change(p, 0);
2764}
2765
2766static void diff_summary(struct diff_filepair *p)
2767{
2768        switch(p->status) {
2769        case DIFF_STATUS_DELETED:
2770                show_file_mode_name("delete", p->one);
2771                break;
2772        case DIFF_STATUS_ADDED:
2773                show_file_mode_name("create", p->two);
2774                break;
2775        case DIFF_STATUS_COPIED:
2776                show_rename_copy("copy", p);
2777                break;
2778        case DIFF_STATUS_RENAMED:
2779                show_rename_copy("rename", p);
2780                break;
2781        default:
2782                if (p->score) {
2783                        fputs(" rewrite ", stdout);
2784                        write_name_quoted(p->two->path, stdout, ' ');
2785                        printf("(%d%%)\n", similarity_index(p));
2786                }
2787                show_mode_change(p, !p->score);
2788                break;
2789        }
2790}
2791
2792struct patch_id_t {
2793        struct xdiff_emit_state xm;
2794        SHA_CTX *ctx;
2795        int patchlen;
2796};
2797
2798static int remove_space(char *line, int len)
2799{
2800        int i;
2801        char *dst = line;
2802        unsigned char c;
2803
2804        for (i = 0; i < len; i++)
2805                if (!isspace((c = line[i])))
2806                        *dst++ = c;
2807
2808        return dst - line;
2809}
2810
2811static void patch_id_consume(void *priv, char *line, unsigned long len)
2812{
2813        struct patch_id_t *data = priv;
2814        int new_len;
2815
2816        /* Ignore line numbers when computing the SHA1 of the patch */
2817        if (!prefixcmp(line, "@@ -"))
2818                return;
2819
2820        new_len = remove_space(line, len);
2821
2822        SHA1_Update(data->ctx, line, new_len);
2823        data->patchlen += new_len;
2824}
2825
2826/* returns 0 upon success, and writes result into sha1 */
2827static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2828{
2829        struct diff_queue_struct *q = &diff_queued_diff;
2830        int i;
2831        SHA_CTX ctx;
2832        struct patch_id_t data;
2833        char buffer[PATH_MAX * 4 + 20];
2834
2835        SHA1_Init(&ctx);
2836        memset(&data, 0, sizeof(struct patch_id_t));
2837        data.ctx = &ctx;
2838        data.xm.consume = patch_id_consume;
2839
2840        for (i = 0; i < q->nr; i++) {
2841                xpparam_t xpp;
2842                xdemitconf_t xecfg;
2843                xdemitcb_t ecb;
2844                mmfile_t mf1, mf2;
2845                struct diff_filepair *p = q->queue[i];
2846                int len1, len2;
2847
2848                memset(&xecfg, 0, sizeof(xecfg));
2849                if (p->status == 0)
2850                        return error("internal diff status error");
2851                if (p->status == DIFF_STATUS_UNKNOWN)
2852                        continue;
2853                if (diff_unmodified_pair(p))
2854                        continue;
2855                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2856                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2857                        continue;
2858                if (DIFF_PAIR_UNMERGED(p))
2859                        continue;
2860
2861                diff_fill_sha1_info(p->one);
2862                diff_fill_sha1_info(p->two);
2863                if (fill_mmfile(&mf1, p->one) < 0 ||
2864                                fill_mmfile(&mf2, p->two) < 0)
2865                        return error("unable to read files to diff");
2866
2867                len1 = remove_space(p->one->path, strlen(p->one->path));
2868                len2 = remove_space(p->two->path, strlen(p->two->path));
2869                if (p->one->mode == 0)
2870                        len1 = snprintf(buffer, sizeof(buffer),
2871                                        "diff--gita/%.*sb/%.*s"
2872                                        "newfilemode%06o"
2873                                        "---/dev/null"
2874                                        "+++b/%.*s",
2875                                        len1, p->one->path,
2876                                        len2, p->two->path,
2877                                        p->two->mode,
2878                                        len2, p->two->path);
2879                else if (p->two->mode == 0)
2880                        len1 = snprintf(buffer, sizeof(buffer),
2881                                        "diff--gita/%.*sb/%.*s"
2882                                        "deletedfilemode%06o"
2883                                        "---a/%.*s"
2884                                        "+++/dev/null",
2885                                        len1, p->one->path,
2886                                        len2, p->two->path,
2887                                        p->one->mode,
2888                                        len1, p->one->path);
2889                else
2890                        len1 = snprintf(buffer, sizeof(buffer),
2891                                        "diff--gita/%.*sb/%.*s"
2892                                        "---a/%.*s"
2893                                        "+++b/%.*s",
2894                                        len1, p->one->path,
2895                                        len2, p->two->path,
2896                                        len1, p->one->path,
2897                                        len2, p->two->path);
2898                SHA1_Update(&ctx, buffer, len1);
2899
2900                xpp.flags = XDF_NEED_MINIMAL;
2901                xecfg.ctxlen = 3;
2902                xecfg.flags = XDL_EMIT_FUNCNAMES;
2903                ecb.outf = xdiff_outf;
2904                ecb.priv = &data;
2905                xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2906        }
2907
2908        SHA1_Final(sha1, &ctx);
2909        return 0;
2910}
2911
2912int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2913{
2914        struct diff_queue_struct *q = &diff_queued_diff;
2915        int i;
2916        int result = diff_get_patch_id(options, sha1);
2917
2918        for (i = 0; i < q->nr; i++)
2919                diff_free_filepair(q->queue[i]);
2920
2921        free(q->queue);
2922        q->queue = NULL;
2923        q->nr = q->alloc = 0;
2924
2925        return result;
2926}
2927
2928static int is_summary_empty(const struct diff_queue_struct *q)
2929{
2930        int i;
2931
2932        for (i = 0; i < q->nr; i++) {
2933                const struct diff_filepair *p = q->queue[i];
2934
2935                switch (p->status) {
2936                case DIFF_STATUS_DELETED:
2937                case DIFF_STATUS_ADDED:
2938                case DIFF_STATUS_COPIED:
2939                case DIFF_STATUS_RENAMED:
2940                        return 0;
2941                default:
2942                        if (p->score)
2943                                return 0;
2944                        if (p->one->mode && p->two->mode &&
2945                            p->one->mode != p->two->mode)
2946                                return 0;
2947                        break;
2948                }
2949        }
2950        return 1;
2951}
2952
2953void diff_flush(struct diff_options *options)
2954{
2955        struct diff_queue_struct *q = &diff_queued_diff;
2956        int i, output_format = options->output_format;
2957        int separator = 0;
2958
2959        /*
2960         * Order: raw, stat, summary, patch
2961         * or:    name/name-status/checkdiff (other bits clear)
2962         */
2963        if (!q->nr)
2964                goto free_queue;
2965
2966        if (output_format & (DIFF_FORMAT_RAW |
2967                             DIFF_FORMAT_NAME |
2968                             DIFF_FORMAT_NAME_STATUS |
2969                             DIFF_FORMAT_CHECKDIFF)) {
2970                for (i = 0; i < q->nr; i++) {
2971                        struct diff_filepair *p = q->queue[i];
2972                        if (check_pair_status(p))
2973                                flush_one_pair(p, options);
2974                }
2975                separator++;
2976        }
2977
2978        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2979                struct diffstat_t diffstat;
2980
2981                memset(&diffstat, 0, sizeof(struct diffstat_t));
2982                diffstat.xm.consume = diffstat_consume;
2983                for (i = 0; i < q->nr; i++) {
2984                        struct diff_filepair *p = q->queue[i];
2985                        if (check_pair_status(p))
2986                                diff_flush_stat(p, options, &diffstat);
2987                }
2988                if (output_format & DIFF_FORMAT_NUMSTAT)
2989                        show_numstat(&diffstat, options);
2990                if (output_format & DIFF_FORMAT_DIFFSTAT)
2991                        show_stats(&diffstat, options);
2992                if (output_format & DIFF_FORMAT_SHORTSTAT)
2993                        show_shortstats(&diffstat);
2994                free_diffstat_info(&diffstat);
2995                separator++;
2996        }
2997
2998        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2999                for (i = 0; i < q->nr; i++)
3000                        diff_summary(q->queue[i]);
3001                separator++;
3002        }
3003
3004        if (output_format & DIFF_FORMAT_PATCH) {
3005                if (separator) {
3006                        if (options->stat_sep) {
3007                                /* attach patch instead of inline */
3008                                fputs(options->stat_sep, stdout);
3009                        } else {
3010                                putchar(options->line_termination);
3011                        }
3012                }
3013
3014                for (i = 0; i < q->nr; i++) {
3015                        struct diff_filepair *p = q->queue[i];
3016                        if (check_pair_status(p))
3017                                diff_flush_patch(p, options);
3018                }
3019        }
3020
3021        if (output_format & DIFF_FORMAT_CALLBACK)
3022                options->format_callback(q, options, options->format_callback_data);
3023
3024        for (i = 0; i < q->nr; i++)
3025                diff_free_filepair(q->queue[i]);
3026free_queue:
3027        free(q->queue);
3028        q->queue = NULL;
3029        q->nr = q->alloc = 0;
3030}
3031
3032static void diffcore_apply_filter(const char *filter)
3033{
3034        int i;
3035        struct diff_queue_struct *q = &diff_queued_diff;
3036        struct diff_queue_struct outq;
3037        outq.queue = NULL;
3038        outq.nr = outq.alloc = 0;
3039
3040        if (!filter)
3041                return;
3042
3043        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3044                int found;
3045                for (i = found = 0; !found && i < q->nr; i++) {
3046                        struct diff_filepair *p = q->queue[i];
3047                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3048                             ((p->score &&
3049                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3050                              (!p->score &&
3051                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3052                            ((p->status != DIFF_STATUS_MODIFIED) &&
3053                             strchr(filter, p->status)))
3054                                found++;
3055                }
3056                if (found)
3057                        return;
3058
3059                /* otherwise we will clear the whole queue
3060                 * by copying the empty outq at the end of this
3061                 * function, but first clear the current entries
3062                 * in the queue.
3063                 */
3064                for (i = 0; i < q->nr; i++)
3065                        diff_free_filepair(q->queue[i]);
3066        }
3067        else {
3068                /* Only the matching ones */
3069                for (i = 0; i < q->nr; i++) {
3070                        struct diff_filepair *p = q->queue[i];
3071
3072                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3073                             ((p->score &&
3074                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3075                              (!p->score &&
3076                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3077                            ((p->status != DIFF_STATUS_MODIFIED) &&
3078                             strchr(filter, p->status)))
3079                                diff_q(&outq, p);
3080                        else
3081                                diff_free_filepair(p);
3082                }
3083        }
3084        free(q->queue);
3085        *q = outq;
3086}
3087
3088/* Check whether two filespecs with the same mode and size are identical */
3089static int diff_filespec_is_identical(struct diff_filespec *one,
3090                                      struct diff_filespec *two)
3091{
3092        if (S_ISGITLINK(one->mode)) {
3093                diff_fill_sha1_info(one);
3094                diff_fill_sha1_info(two);
3095                return !hashcmp(one->sha1, two->sha1);
3096        }
3097        if (diff_populate_filespec(one, 0))
3098                return 0;
3099        if (diff_populate_filespec(two, 0))
3100                return 0;
3101        return !memcmp(one->data, two->data, one->size);
3102}
3103
3104static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3105{
3106        int i;
3107        struct diff_queue_struct *q = &diff_queued_diff;
3108        struct diff_queue_struct outq;
3109        outq.queue = NULL;
3110        outq.nr = outq.alloc = 0;
3111
3112        for (i = 0; i < q->nr; i++) {
3113                struct diff_filepair *p = q->queue[i];
3114
3115                /*
3116                 * 1. Entries that come from stat info dirtyness
3117                 *    always have both sides (iow, not create/delete),
3118                 *    one side of the object name is unknown, with
3119                 *    the same mode and size.  Keep the ones that
3120                 *    do not match these criteria.  They have real
3121                 *    differences.
3122                 *
3123                 * 2. At this point, the file is known to be modified,
3124                 *    with the same mode and size, and the object
3125                 *    name of one side is unknown.  Need to inspect
3126                 *    the identical contents.
3127                 */
3128                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3129                    !DIFF_FILE_VALID(p->two) ||
3130                    (p->one->sha1_valid && p->two->sha1_valid) ||
3131                    (p->one->mode != p->two->mode) ||
3132                    diff_populate_filespec(p->one, 1) ||
3133                    diff_populate_filespec(p->two, 1) ||
3134                    (p->one->size != p->two->size) ||
3135                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3136                        diff_q(&outq, p);
3137                else {
3138                        /*
3139                         * The caller can subtract 1 from skip_stat_unmatch
3140                         * to determine how many paths were dirty only
3141                         * due to stat info mismatch.
3142                         */
3143                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3144                                diffopt->skip_stat_unmatch++;
3145                        diff_free_filepair(p);
3146                }
3147        }
3148        free(q->queue);
3149        *q = outq;
3150}
3151
3152void diffcore_std(struct diff_options *options)
3153{
3154        if (DIFF_OPT_TST(options, QUIET))
3155                return;
3156
3157        if (options->skip_stat_unmatch && !DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3158                diffcore_skip_stat_unmatch(options);
3159        if (options->break_opt != -1)
3160                diffcore_break(options->break_opt);
3161        if (options->detect_rename)
3162                diffcore_rename(options);
3163        if (options->break_opt != -1)
3164                diffcore_merge_broken();
3165        if (options->pickaxe)
3166                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3167        if (options->orderfile)
3168                diffcore_order(options->orderfile);
3169        diff_resolve_rename_copy();
3170        diffcore_apply_filter(options->filter);
3171
3172        if (diff_queued_diff.nr)
3173                DIFF_OPT_SET(options, HAS_CHANGES);
3174        else
3175                DIFF_OPT_CLR(options, HAS_CHANGES);
3176}
3177
3178int diff_result_code(struct diff_options *opt, int status)
3179{
3180        int result = 0;
3181        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3182            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3183                return status;
3184        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3185            DIFF_OPT_TST(opt, HAS_CHANGES))
3186                result |= 01;
3187        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3188            DIFF_OPT_TST(opt, CHECK_FAILED))
3189                result |= 02;
3190        return result;
3191}
3192
3193void diff_addremove(struct diff_options *options,
3194                    int addremove, unsigned mode,
3195                    const unsigned char *sha1,
3196                    const char *base, const char *path)
3197{
3198        char concatpath[PATH_MAX];
3199        struct diff_filespec *one, *two;
3200
3201        /* This may look odd, but it is a preparation for
3202         * feeding "there are unchanged files which should
3203         * not produce diffs, but when you are doing copy
3204         * detection you would need them, so here they are"
3205         * entries to the diff-core.  They will be prefixed
3206         * with something like '=' or '*' (I haven't decided
3207         * which but should not make any difference).
3208         * Feeding the same new and old to diff_change()
3209         * also has the same effect.
3210         * Before the final output happens, they are pruned after
3211         * merged into rename/copy pairs as appropriate.
3212         */
3213        if (DIFF_OPT_TST(options, REVERSE_DIFF))
3214                addremove = (addremove == '+' ? '-' :
3215                             addremove == '-' ? '+' : addremove);
3216
3217        if (!path) path = "";
3218        sprintf(concatpath, "%s%s", base, path);
3219        one = alloc_filespec(concatpath);
3220        two = alloc_filespec(concatpath);
3221
3222        if (addremove != '+')
3223                fill_filespec(one, sha1, mode);
3224        if (addremove != '-')
3225                fill_filespec(two, sha1, mode);
3226
3227        diff_queue(&diff_queued_diff, one, two);
3228        DIFF_OPT_SET(options, HAS_CHANGES);
3229}
3230
3231void diff_change(struct diff_options *options,
3232                 unsigned old_mode, unsigned new_mode,
3233                 const unsigned char *old_sha1,
3234                 const unsigned char *new_sha1,
3235                 const char *base, const char *path)
3236{
3237        char concatpath[PATH_MAX];
3238        struct diff_filespec *one, *two;
3239
3240        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3241                unsigned tmp;
3242                const unsigned char *tmp_c;
3243                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3244                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3245        }
3246        if (!path) path = "";
3247        sprintf(concatpath, "%s%s", base, path);
3248        one = alloc_filespec(concatpath);
3249        two = alloc_filespec(concatpath);
3250        fill_filespec(one, old_sha1, old_mode);
3251        fill_filespec(two, new_sha1, new_mode);
3252
3253        diff_queue(&diff_queued_diff, one, two);
3254        DIFF_OPT_SET(options, HAS_CHANGES);
3255}
3256
3257void diff_unmerge(struct diff_options *options,
3258                  const char *path,
3259                  unsigned mode, const unsigned char *sha1)
3260{
3261        struct diff_filespec *one, *two;
3262        one = alloc_filespec(path);
3263        two = alloc_filespec(path);
3264        fill_filespec(one, sha1, mode);
3265        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3266}