diff.con commit core.whitespace: add test for diff whitespace error highlighting (49e703a)
   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);
 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        xdl_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        const char **label_path;
 458        struct diff_words_data *diff_words;
 459        int *found_changesp;
 460};
 461
 462static void free_diff_words_data(struct emit_callback *ecbdata)
 463{
 464        if (ecbdata->diff_words) {
 465                /* flush buffers */
 466                if (ecbdata->diff_words->minus.text.size ||
 467                                ecbdata->diff_words->plus.text.size)
 468                        diff_words_show(ecbdata->diff_words);
 469
 470                if (ecbdata->diff_words->minus.text.ptr)
 471                        free (ecbdata->diff_words->minus.text.ptr);
 472                if (ecbdata->diff_words->plus.text.ptr)
 473                        free (ecbdata->diff_words->plus.text.ptr);
 474                free(ecbdata->diff_words);
 475                ecbdata->diff_words = NULL;
 476        }
 477}
 478
 479const char *diff_get_color(int diff_use_color, enum color_diff ix)
 480{
 481        if (diff_use_color)
 482                return diff_colors[ix];
 483        return "";
 484}
 485
 486static void emit_line(const char *set, const char *reset, const char *line, int len)
 487{
 488        if (len > 0 && line[len-1] == '\n')
 489                len--;
 490        fputs(set, stdout);
 491        fwrite(line, len, 1, stdout);
 492        puts(reset);
 493}
 494
 495static void emit_line_with_ws(int nparents,
 496                const char *set, const char *reset, const char *ws,
 497                const char *line, int len)
 498{
 499        int col0 = nparents;
 500        int last_tab_in_indent = -1;
 501        int last_space_in_indent = -1;
 502        int i;
 503        int tail = len;
 504        int need_highlight_leading_space = 0;
 505        /*
 506         * The line is a newly added line.  Does it have funny leading
 507         * whitespaces?  In indent, SP should never precede a TAB.  In
 508         * addition, under "indent with non tab" rule, there should not
 509         * be more than 8 consecutive spaces.
 510         */
 511        for (i = col0; i < len; i++) {
 512                if (line[i] == '\t') {
 513                        last_tab_in_indent = i;
 514                        if ((whitespace_rule & WS_SPACE_BEFORE_TAB) &&
 515                            0 <= last_space_in_indent)
 516                                need_highlight_leading_space = 1;
 517                }
 518                else if (line[i] == ' ')
 519                        last_space_in_indent = i;
 520                else
 521                        break;
 522        }
 523        if ((whitespace_rule & WS_INDENT_WITH_NON_TAB) &&
 524            0 <= last_space_in_indent &&
 525            last_tab_in_indent < 0 &&
 526            8 <= (i - col0)) {
 527                last_tab_in_indent = i;
 528                need_highlight_leading_space = 1;
 529        }
 530        fputs(set, stdout);
 531        fwrite(line, col0, 1, stdout);
 532        fputs(reset, stdout);
 533        if (((i == len) || line[i] == '\n') && i != col0) {
 534                /* The whole line was indent */
 535                emit_line(ws, reset, line + col0, len - col0);
 536                return;
 537        }
 538        i = col0;
 539        if (need_highlight_leading_space) {
 540                while (i < last_tab_in_indent) {
 541                        if (line[i] == ' ') {
 542                                fputs(ws, stdout);
 543                                putchar(' ');
 544                                fputs(reset, stdout);
 545                        }
 546                        else
 547                                putchar(line[i]);
 548                        i++;
 549                }
 550        }
 551        tail = len - 1;
 552        if (line[tail] == '\n' && i < tail)
 553                tail--;
 554        if (whitespace_rule & WS_TRAILING_SPACE) {
 555                while (i < tail) {
 556                        if (!isspace(line[tail]))
 557                                break;
 558                        tail--;
 559                }
 560        }
 561        if ((i < tail && line[tail + 1] != '\n')) {
 562                /* This has whitespace between tail+1..len */
 563                fputs(set, stdout);
 564                fwrite(line + i, tail - i + 1, 1, stdout);
 565                fputs(reset, stdout);
 566                emit_line(ws, reset, line + tail + 1, len - tail - 1);
 567        }
 568        else
 569                emit_line(set, reset, line + i, len - i);
 570}
 571
 572static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 573{
 574        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 575        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 576
 577        if (!*ws)
 578                emit_line(set, reset, line, len);
 579        else
 580                emit_line_with_ws(ecbdata->nparents, set, reset, ws,
 581                                line, len);
 582}
 583
 584static void fn_out_consume(void *priv, char *line, unsigned long len)
 585{
 586        int i;
 587        int color;
 588        struct emit_callback *ecbdata = priv;
 589        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 590        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 591
 592        *(ecbdata->found_changesp) = 1;
 593
 594        if (ecbdata->label_path[0]) {
 595                const char *name_a_tab, *name_b_tab;
 596
 597                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 598                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 599
 600                printf("%s--- %s%s%s\n",
 601                       set, ecbdata->label_path[0], reset, name_a_tab);
 602                printf("%s+++ %s%s%s\n",
 603                       set, ecbdata->label_path[1], reset, name_b_tab);
 604                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 605        }
 606
 607        /* This is not really necessary for now because
 608         * this codepath only deals with two-way diffs.
 609         */
 610        for (i = 0; i < len && line[i] == '@'; i++)
 611                ;
 612        if (2 <= i && i < len && line[i] == ' ') {
 613                ecbdata->nparents = i - 1;
 614                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 615                          reset, line, len);
 616                return;
 617        }
 618
 619        if (len < ecbdata->nparents) {
 620                set = reset;
 621                emit_line(reset, reset, line, len);
 622                return;
 623        }
 624
 625        color = DIFF_PLAIN;
 626        if (ecbdata->diff_words && ecbdata->nparents != 1)
 627                /* fall back to normal diff */
 628                free_diff_words_data(ecbdata);
 629        if (ecbdata->diff_words) {
 630                if (line[0] == '-') {
 631                        diff_words_append(line, len,
 632                                          &ecbdata->diff_words->minus);
 633                        return;
 634                } else if (line[0] == '+') {
 635                        diff_words_append(line, len,
 636                                          &ecbdata->diff_words->plus);
 637                        return;
 638                }
 639                if (ecbdata->diff_words->minus.text.size ||
 640                    ecbdata->diff_words->plus.text.size)
 641                        diff_words_show(ecbdata->diff_words);
 642                line++;
 643                len--;
 644                emit_line(set, reset, line, len);
 645                return;
 646        }
 647        for (i = 0; i < ecbdata->nparents && len; i++) {
 648                if (line[i] == '-')
 649                        color = DIFF_FILE_OLD;
 650                else if (line[i] == '+')
 651                        color = DIFF_FILE_NEW;
 652        }
 653
 654        if (color != DIFF_FILE_NEW) {
 655                emit_line(diff_get_color(ecbdata->color_diff, color),
 656                          reset, line, len);
 657                return;
 658        }
 659        emit_add_line(reset, ecbdata, line, len);
 660}
 661
 662static char *pprint_rename(const char *a, const char *b)
 663{
 664        const char *old = a;
 665        const char *new = b;
 666        struct strbuf name;
 667        int pfx_length, sfx_length;
 668        int len_a = strlen(a);
 669        int len_b = strlen(b);
 670        int a_midlen, b_midlen;
 671        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 672        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 673
 674        strbuf_init(&name, 0);
 675        if (qlen_a || qlen_b) {
 676                quote_c_style(a, &name, NULL, 0);
 677                strbuf_addstr(&name, " => ");
 678                quote_c_style(b, &name, NULL, 0);
 679                return strbuf_detach(&name, NULL);
 680        }
 681
 682        /* Find common prefix */
 683        pfx_length = 0;
 684        while (*old && *new && *old == *new) {
 685                if (*old == '/')
 686                        pfx_length = old - a + 1;
 687                old++;
 688                new++;
 689        }
 690
 691        /* Find common suffix */
 692        old = a + len_a;
 693        new = b + len_b;
 694        sfx_length = 0;
 695        while (a <= old && b <= new && *old == *new) {
 696                if (*old == '/')
 697                        sfx_length = len_a - (old - a);
 698                old--;
 699                new--;
 700        }
 701
 702        /*
 703         * pfx{mid-a => mid-b}sfx
 704         * {pfx-a => pfx-b}sfx
 705         * pfx{sfx-a => sfx-b}
 706         * name-a => name-b
 707         */
 708        a_midlen = len_a - pfx_length - sfx_length;
 709        b_midlen = len_b - pfx_length - sfx_length;
 710        if (a_midlen < 0)
 711                a_midlen = 0;
 712        if (b_midlen < 0)
 713                b_midlen = 0;
 714
 715        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 716        if (pfx_length + sfx_length) {
 717                strbuf_add(&name, a, pfx_length);
 718                strbuf_addch(&name, '{');
 719        }
 720        strbuf_add(&name, a + pfx_length, a_midlen);
 721        strbuf_addstr(&name, " => ");
 722        strbuf_add(&name, b + pfx_length, b_midlen);
 723        if (pfx_length + sfx_length) {
 724                strbuf_addch(&name, '}');
 725                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 726        }
 727        return strbuf_detach(&name, NULL);
 728}
 729
 730struct diffstat_t {
 731        struct xdiff_emit_state xm;
 732
 733        int nr;
 734        int alloc;
 735        struct diffstat_file {
 736                char *name;
 737                unsigned is_unmerged:1;
 738                unsigned is_binary:1;
 739                unsigned is_renamed:1;
 740                unsigned int added, deleted;
 741        } **files;
 742};
 743
 744static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 745                                          const char *name_a,
 746                                          const char *name_b)
 747{
 748        struct diffstat_file *x;
 749        x = xcalloc(sizeof (*x), 1);
 750        if (diffstat->nr == diffstat->alloc) {
 751                diffstat->alloc = alloc_nr(diffstat->alloc);
 752                diffstat->files = xrealloc(diffstat->files,
 753                                diffstat->alloc * sizeof(x));
 754        }
 755        diffstat->files[diffstat->nr++] = x;
 756        if (name_b) {
 757                x->name = pprint_rename(name_a, name_b);
 758                x->is_renamed = 1;
 759        }
 760        else
 761                x->name = xstrdup(name_a);
 762        return x;
 763}
 764
 765static void diffstat_consume(void *priv, char *line, unsigned long len)
 766{
 767        struct diffstat_t *diffstat = priv;
 768        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 769
 770        if (line[0] == '+')
 771                x->added++;
 772        else if (line[0] == '-')
 773                x->deleted++;
 774}
 775
 776const char mime_boundary_leader[] = "------------";
 777
 778static int scale_linear(int it, int width, int max_change)
 779{
 780        /*
 781         * make sure that at least one '-' is printed if there were deletions,
 782         * and likewise for '+'.
 783         */
 784        if (max_change < 2)
 785                return it;
 786        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 787}
 788
 789static void show_name(const char *prefix, const char *name, int len,
 790                      const char *reset, const char *set)
 791{
 792        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 793}
 794
 795static void show_graph(char ch, int cnt, const char *set, const char *reset)
 796{
 797        if (cnt <= 0)
 798                return;
 799        printf("%s", set);
 800        while (cnt--)
 801                putchar(ch);
 802        printf("%s", reset);
 803}
 804
 805static void show_stats(struct diffstat_t* data, struct diff_options *options)
 806{
 807        int i, len, add, del, total, adds = 0, dels = 0;
 808        int max_change = 0, max_len = 0;
 809        int total_files = data->nr;
 810        int width, name_width;
 811        const char *reset, *set, *add_c, *del_c;
 812
 813        if (data->nr == 0)
 814                return;
 815
 816        width = options->stat_width ? options->stat_width : 80;
 817        name_width = options->stat_name_width ? options->stat_name_width : 50;
 818
 819        /* Sanity: give at least 5 columns to the graph,
 820         * but leave at least 10 columns for the name.
 821         */
 822        if (width < name_width + 15) {
 823                if (name_width <= 25)
 824                        width = name_width + 15;
 825                else
 826                        name_width = width - 15;
 827        }
 828
 829        /* Find the longest filename and max number of changes */
 830        reset = diff_get_color(options->color_diff, DIFF_RESET);
 831        set = diff_get_color(options->color_diff, DIFF_PLAIN);
 832        add_c = diff_get_color(options->color_diff, DIFF_FILE_NEW);
 833        del_c = diff_get_color(options->color_diff, DIFF_FILE_OLD);
 834
 835        for (i = 0; i < data->nr; i++) {
 836                struct diffstat_file *file = data->files[i];
 837                int change = file->added + file->deleted;
 838
 839                if (!file->is_renamed) {  /* renames are already quoted by pprint_rename */
 840                        struct strbuf buf;
 841                        strbuf_init(&buf, 0);
 842                        if (quote_c_style(file->name, &buf, NULL, 0)) {
 843                                free(file->name);
 844                                file->name = strbuf_detach(&buf, NULL);
 845                        } else {
 846                                strbuf_release(&buf);
 847                        }
 848                }
 849
 850                len = strlen(file->name);
 851                if (max_len < len)
 852                        max_len = len;
 853
 854                if (file->is_binary || file->is_unmerged)
 855                        continue;
 856                if (max_change < change)
 857                        max_change = change;
 858        }
 859
 860        /* Compute the width of the graph part;
 861         * 10 is for one blank at the beginning of the line plus
 862         * " | count " between the name and the graph.
 863         *
 864         * From here on, name_width is the width of the name area,
 865         * and width is the width of the graph area.
 866         */
 867        name_width = (name_width < max_len) ? name_width : max_len;
 868        if (width < (name_width + 10) + max_change)
 869                width = width - (name_width + 10);
 870        else
 871                width = max_change;
 872
 873        for (i = 0; i < data->nr; i++) {
 874                const char *prefix = "";
 875                char *name = data->files[i]->name;
 876                int added = data->files[i]->added;
 877                int deleted = data->files[i]->deleted;
 878                int name_len;
 879
 880                /*
 881                 * "scale" the filename
 882                 */
 883                len = name_width;
 884                name_len = strlen(name);
 885                if (name_width < name_len) {
 886                        char *slash;
 887                        prefix = "...";
 888                        len -= 3;
 889                        name += name_len - len;
 890                        slash = strchr(name, '/');
 891                        if (slash)
 892                                name = slash;
 893                }
 894
 895                if (data->files[i]->is_binary) {
 896                        show_name(prefix, name, len, reset, set);
 897                        printf("  Bin ");
 898                        printf("%s%d%s", del_c, deleted, reset);
 899                        printf(" -> ");
 900                        printf("%s%d%s", add_c, added, reset);
 901                        printf(" bytes");
 902                        printf("\n");
 903                        goto free_diffstat_file;
 904                }
 905                else if (data->files[i]->is_unmerged) {
 906                        show_name(prefix, name, len, reset, set);
 907                        printf("  Unmerged\n");
 908                        goto free_diffstat_file;
 909                }
 910                else if (!data->files[i]->is_renamed &&
 911                         (added + deleted == 0)) {
 912                        total_files--;
 913                        goto free_diffstat_file;
 914                }
 915
 916                /*
 917                 * scale the add/delete
 918                 */
 919                add = added;
 920                del = deleted;
 921                total = add + del;
 922                adds += add;
 923                dels += del;
 924
 925                if (width <= max_change) {
 926                        add = scale_linear(add, width, max_change);
 927                        del = scale_linear(del, width, max_change);
 928                        total = add + del;
 929                }
 930                show_name(prefix, name, len, reset, set);
 931                printf("%5d ", added + deleted);
 932                show_graph('+', add, add_c, reset);
 933                show_graph('-', del, del_c, reset);
 934                putchar('\n');
 935        free_diffstat_file:
 936                free(data->files[i]->name);
 937                free(data->files[i]);
 938        }
 939        free(data->files);
 940        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 941               set, total_files, adds, dels, reset);
 942}
 943
 944static void show_shortstats(struct diffstat_t* data)
 945{
 946        int i, adds = 0, dels = 0, total_files = data->nr;
 947
 948        if (data->nr == 0)
 949                return;
 950
 951        for (i = 0; i < data->nr; i++) {
 952                if (!data->files[i]->is_binary &&
 953                    !data->files[i]->is_unmerged) {
 954                        int added = data->files[i]->added;
 955                        int deleted= data->files[i]->deleted;
 956                        if (!data->files[i]->is_renamed &&
 957                            (added + deleted == 0)) {
 958                                total_files--;
 959                        } else {
 960                                adds += added;
 961                                dels += deleted;
 962                        }
 963                }
 964                free(data->files[i]->name);
 965                free(data->files[i]);
 966        }
 967        free(data->files);
 968
 969        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 970               total_files, adds, dels);
 971}
 972
 973static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 974{
 975        int i;
 976
 977        for (i = 0; i < data->nr; i++) {
 978                struct diffstat_file *file = data->files[i];
 979
 980                if (file->is_binary)
 981                        printf("-\t-\t");
 982                else
 983                        printf("%d\t%d\t", file->added, file->deleted);
 984                if (!file->is_renamed) {
 985                        write_name_quoted(file->name, stdout, options->line_termination);
 986                } else {
 987                        fputs(file->name, stdout);
 988                        putchar(options->line_termination);
 989                }
 990        }
 991}
 992
 993struct checkdiff_t {
 994        struct xdiff_emit_state xm;
 995        const char *filename;
 996        int lineno, color_diff;
 997};
 998
 999static void checkdiff_consume(void *priv, char *line, unsigned long len)
1000{
1001        struct checkdiff_t *data = priv;
1002        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
1003        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
1004        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
1005
1006        if (line[0] == '+') {
1007                int i, spaces = 0, space_before_tab = 0, white_space_at_end = 0;
1008
1009                /* check space before tab */
1010                for (i = 1; i < len && (line[i] == ' ' || line[i] == '\t'); i++)
1011                        if (line[i] == ' ')
1012                                spaces++;
1013                if (line[i - 1] == '\t' && spaces)
1014                        space_before_tab = 1;
1015
1016                /* check white space at line end */
1017                if (line[len - 1] == '\n')
1018                        len--;
1019                if (isspace(line[len - 1]))
1020                        white_space_at_end = 1;
1021
1022                if (space_before_tab || white_space_at_end) {
1023                        printf("%s:%d: %s", data->filename, data->lineno, ws);
1024                        if (space_before_tab) {
1025                                printf("space before tab");
1026                                if (white_space_at_end)
1027                                        putchar(',');
1028                        }
1029                        if (white_space_at_end)
1030                                printf("white space at end");
1031                        printf(":%s ", reset);
1032                        emit_line_with_ws(1, set, reset, ws, line, len);
1033                }
1034
1035                data->lineno++;
1036        } else if (line[0] == ' ')
1037                data->lineno++;
1038        else if (line[0] == '@') {
1039                char *plus = strchr(line, '+');
1040                if (plus)
1041                        data->lineno = strtol(plus, NULL, 10);
1042                else
1043                        die("invalid diff");
1044        }
1045}
1046
1047static unsigned char *deflate_it(char *data,
1048                                 unsigned long size,
1049                                 unsigned long *result_size)
1050{
1051        int bound;
1052        unsigned char *deflated;
1053        z_stream stream;
1054
1055        memset(&stream, 0, sizeof(stream));
1056        deflateInit(&stream, zlib_compression_level);
1057        bound = deflateBound(&stream, size);
1058        deflated = xmalloc(bound);
1059        stream.next_out = deflated;
1060        stream.avail_out = bound;
1061
1062        stream.next_in = (unsigned char *)data;
1063        stream.avail_in = size;
1064        while (deflate(&stream, Z_FINISH) == Z_OK)
1065                ; /* nothing */
1066        deflateEnd(&stream);
1067        *result_size = stream.total_out;
1068        return deflated;
1069}
1070
1071static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1072{
1073        void *cp;
1074        void *delta;
1075        void *deflated;
1076        void *data;
1077        unsigned long orig_size;
1078        unsigned long delta_size;
1079        unsigned long deflate_size;
1080        unsigned long data_size;
1081
1082        /* We could do deflated delta, or we could do just deflated two,
1083         * whichever is smaller.
1084         */
1085        delta = NULL;
1086        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1087        if (one->size && two->size) {
1088                delta = diff_delta(one->ptr, one->size,
1089                                   two->ptr, two->size,
1090                                   &delta_size, deflate_size);
1091                if (delta) {
1092                        void *to_free = delta;
1093                        orig_size = delta_size;
1094                        delta = deflate_it(delta, delta_size, &delta_size);
1095                        free(to_free);
1096                }
1097        }
1098
1099        if (delta && delta_size < deflate_size) {
1100                printf("delta %lu\n", orig_size);
1101                free(deflated);
1102                data = delta;
1103                data_size = delta_size;
1104        }
1105        else {
1106                printf("literal %lu\n", two->size);
1107                free(delta);
1108                data = deflated;
1109                data_size = deflate_size;
1110        }
1111
1112        /* emit data encoded in base85 */
1113        cp = data;
1114        while (data_size) {
1115                int bytes = (52 < data_size) ? 52 : data_size;
1116                char line[70];
1117                data_size -= bytes;
1118                if (bytes <= 26)
1119                        line[0] = bytes + 'A' - 1;
1120                else
1121                        line[0] = bytes - 26 + 'a' - 1;
1122                encode_85(line + 1, cp, bytes);
1123                cp = (char *) cp + bytes;
1124                puts(line);
1125        }
1126        printf("\n");
1127        free(data);
1128}
1129
1130static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1131{
1132        printf("GIT binary patch\n");
1133        emit_binary_diff_body(one, two);
1134        emit_binary_diff_body(two, one);
1135}
1136
1137static void setup_diff_attr_check(struct git_attr_check *check)
1138{
1139        static struct git_attr *attr_diff;
1140
1141        if (!attr_diff) {
1142                attr_diff = git_attr("diff", 4);
1143        }
1144        check[0].attr = attr_diff;
1145}
1146
1147static void diff_filespec_check_attr(struct diff_filespec *one)
1148{
1149        struct git_attr_check attr_diff_check;
1150        int check_from_data = 0;
1151
1152        if (one->checked_attr)
1153                return;
1154
1155        setup_diff_attr_check(&attr_diff_check);
1156        one->is_binary = 0;
1157        one->funcname_pattern_ident = NULL;
1158
1159        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1160                const char *value;
1161
1162                /* binaryness */
1163                value = attr_diff_check.value;
1164                if (ATTR_TRUE(value))
1165                        ;
1166                else if (ATTR_FALSE(value))
1167                        one->is_binary = 1;
1168                else
1169                        check_from_data = 1;
1170
1171                /* funcname pattern ident */
1172                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1173                        ;
1174                else
1175                        one->funcname_pattern_ident = value;
1176        }
1177
1178        if (check_from_data) {
1179                if (!one->data && DIFF_FILE_VALID(one))
1180                        diff_populate_filespec(one, 0);
1181
1182                if (one->data)
1183                        one->is_binary = buffer_is_binary(one->data, one->size);
1184        }
1185}
1186
1187int diff_filespec_is_binary(struct diff_filespec *one)
1188{
1189        diff_filespec_check_attr(one);
1190        return one->is_binary;
1191}
1192
1193static const char *funcname_pattern(const char *ident)
1194{
1195        struct funcname_pattern *pp;
1196
1197        read_config_if_needed();
1198        for (pp = funcname_pattern_list; pp; pp = pp->next)
1199                if (!strcmp(ident, pp->name))
1200                        return pp->pattern;
1201        return NULL;
1202}
1203
1204static struct builtin_funcname_pattern {
1205        const char *name;
1206        const char *pattern;
1207} builtin_funcname_pattern[] = {
1208        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1209                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1210                        "^[     ]*\\(\\([       ]*"
1211                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1212                        "[      ]*([^;]*$\\)" },
1213        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1214};
1215
1216static const char *diff_funcname_pattern(struct diff_filespec *one)
1217{
1218        const char *ident, *pattern;
1219        int i;
1220
1221        diff_filespec_check_attr(one);
1222        ident = one->funcname_pattern_ident;
1223
1224        if (!ident)
1225                /*
1226                 * If the config file has "funcname.default" defined, that
1227                 * regexp is used; otherwise NULL is returned and xemit uses
1228                 * the built-in default.
1229                 */
1230                return funcname_pattern("default");
1231
1232        /* Look up custom "funcname.$ident" regexp from config. */
1233        pattern = funcname_pattern(ident);
1234        if (pattern)
1235                return pattern;
1236
1237        /*
1238         * And define built-in fallback patterns here.  Note that
1239         * these can be overriden by the user's config settings.
1240         */
1241        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1242                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1243                        return builtin_funcname_pattern[i].pattern;
1244
1245        return NULL;
1246}
1247
1248static void builtin_diff(const char *name_a,
1249                         const char *name_b,
1250                         struct diff_filespec *one,
1251                         struct diff_filespec *two,
1252                         const char *xfrm_msg,
1253                         struct diff_options *o,
1254                         int complete_rewrite)
1255{
1256        mmfile_t mf1, mf2;
1257        const char *lbl[2];
1258        char *a_one, *b_two;
1259        const char *set = diff_get_color(o->color_diff, DIFF_METAINFO);
1260        const char *reset = diff_get_color(o->color_diff, DIFF_RESET);
1261
1262        a_one = quote_two("a/", name_a + (*name_a == '/'));
1263        b_two = quote_two("b/", name_b + (*name_b == '/'));
1264        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1265        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1266        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1267        if (lbl[0][0] == '/') {
1268                /* /dev/null */
1269                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1270                if (xfrm_msg && xfrm_msg[0])
1271                        printf("%s%s%s\n", set, xfrm_msg, reset);
1272        }
1273        else if (lbl[1][0] == '/') {
1274                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1275                if (xfrm_msg && xfrm_msg[0])
1276                        printf("%s%s%s\n", set, xfrm_msg, reset);
1277        }
1278        else {
1279                if (one->mode != two->mode) {
1280                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1281                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1282                }
1283                if (xfrm_msg && xfrm_msg[0])
1284                        printf("%s%s%s\n", set, xfrm_msg, reset);
1285                /*
1286                 * we do not run diff between different kind
1287                 * of objects.
1288                 */
1289                if ((one->mode ^ two->mode) & S_IFMT)
1290                        goto free_ab_and_return;
1291                if (complete_rewrite) {
1292                        emit_rewrite_diff(name_a, name_b, one, two,
1293                                        o->color_diff);
1294                        o->found_changes = 1;
1295                        goto free_ab_and_return;
1296                }
1297        }
1298
1299        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1300                die("unable to read files to diff");
1301
1302        if (!o->text &&
1303            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1304                /* Quite common confusing case */
1305                if (mf1.size == mf2.size &&
1306                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1307                        goto free_ab_and_return;
1308                if (o->binary)
1309                        emit_binary_diff(&mf1, &mf2);
1310                else
1311                        printf("Binary files %s and %s differ\n",
1312                               lbl[0], lbl[1]);
1313                o->found_changes = 1;
1314        }
1315        else {
1316                /* Crazy xdl interfaces.. */
1317                const char *diffopts = getenv("GIT_DIFF_OPTS");
1318                xpparam_t xpp;
1319                xdemitconf_t xecfg;
1320                xdemitcb_t ecb;
1321                struct emit_callback ecbdata;
1322                const char *funcname_pattern;
1323
1324                funcname_pattern = diff_funcname_pattern(one);
1325                if (!funcname_pattern)
1326                        funcname_pattern = diff_funcname_pattern(two);
1327
1328                memset(&xecfg, 0, sizeof(xecfg));
1329                memset(&ecbdata, 0, sizeof(ecbdata));
1330                ecbdata.label_path = lbl;
1331                ecbdata.color_diff = o->color_diff;
1332                ecbdata.found_changesp = &o->found_changes;
1333                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1334                xecfg.ctxlen = o->context;
1335                xecfg.flags = XDL_EMIT_FUNCNAMES;
1336                if (funcname_pattern)
1337                        xdiff_set_find_func(&xecfg, funcname_pattern);
1338                if (!diffopts)
1339                        ;
1340                else if (!prefixcmp(diffopts, "--unified="))
1341                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1342                else if (!prefixcmp(diffopts, "-u"))
1343                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1344                ecb.outf = xdiff_outf;
1345                ecb.priv = &ecbdata;
1346                ecbdata.xm.consume = fn_out_consume;
1347                if (o->color_diff_words)
1348                        ecbdata.diff_words =
1349                                xcalloc(1, sizeof(struct diff_words_data));
1350                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1351                if (o->color_diff_words)
1352                        free_diff_words_data(&ecbdata);
1353        }
1354
1355 free_ab_and_return:
1356        diff_free_filespec_data(one);
1357        diff_free_filespec_data(two);
1358        free(a_one);
1359        free(b_two);
1360        return;
1361}
1362
1363static void builtin_diffstat(const char *name_a, const char *name_b,
1364                             struct diff_filespec *one,
1365                             struct diff_filespec *two,
1366                             struct diffstat_t *diffstat,
1367                             struct diff_options *o,
1368                             int complete_rewrite)
1369{
1370        mmfile_t mf1, mf2;
1371        struct diffstat_file *data;
1372
1373        data = diffstat_add(diffstat, name_a, name_b);
1374
1375        if (!one || !two) {
1376                data->is_unmerged = 1;
1377                return;
1378        }
1379        if (complete_rewrite) {
1380                diff_populate_filespec(one, 0);
1381                diff_populate_filespec(two, 0);
1382                data->deleted = count_lines(one->data, one->size);
1383                data->added = count_lines(two->data, two->size);
1384                goto free_and_return;
1385        }
1386        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1387                die("unable to read files to diff");
1388
1389        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1390                data->is_binary = 1;
1391                data->added = mf2.size;
1392                data->deleted = mf1.size;
1393        } else {
1394                /* Crazy xdl interfaces.. */
1395                xpparam_t xpp;
1396                xdemitconf_t xecfg;
1397                xdemitcb_t ecb;
1398
1399                memset(&xecfg, 0, sizeof(xecfg));
1400                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1401                ecb.outf = xdiff_outf;
1402                ecb.priv = diffstat;
1403                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1404        }
1405
1406 free_and_return:
1407        diff_free_filespec_data(one);
1408        diff_free_filespec_data(two);
1409}
1410
1411static void builtin_checkdiff(const char *name_a, const char *name_b,
1412                             struct diff_filespec *one,
1413                             struct diff_filespec *two, struct diff_options *o)
1414{
1415        mmfile_t mf1, mf2;
1416        struct checkdiff_t data;
1417
1418        if (!two)
1419                return;
1420
1421        memset(&data, 0, sizeof(data));
1422        data.xm.consume = checkdiff_consume;
1423        data.filename = name_b ? name_b : name_a;
1424        data.lineno = 0;
1425        data.color_diff = o->color_diff;
1426
1427        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1428                die("unable to read files to diff");
1429
1430        if (diff_filespec_is_binary(two))
1431                goto free_and_return;
1432        else {
1433                /* Crazy xdl interfaces.. */
1434                xpparam_t xpp;
1435                xdemitconf_t xecfg;
1436                xdemitcb_t ecb;
1437
1438                memset(&xecfg, 0, sizeof(xecfg));
1439                xpp.flags = XDF_NEED_MINIMAL;
1440                ecb.outf = xdiff_outf;
1441                ecb.priv = &data;
1442                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1443        }
1444 free_and_return:
1445        diff_free_filespec_data(one);
1446        diff_free_filespec_data(two);
1447}
1448
1449struct diff_filespec *alloc_filespec(const char *path)
1450{
1451        int namelen = strlen(path);
1452        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1453
1454        memset(spec, 0, sizeof(*spec));
1455        spec->path = (char *)(spec + 1);
1456        memcpy(spec->path, path, namelen+1);
1457        spec->count = 1;
1458        return spec;
1459}
1460
1461void free_filespec(struct diff_filespec *spec)
1462{
1463        if (!--spec->count) {
1464                diff_free_filespec_data(spec);
1465                free(spec);
1466        }
1467}
1468
1469void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1470                   unsigned short mode)
1471{
1472        if (mode) {
1473                spec->mode = canon_mode(mode);
1474                hashcpy(spec->sha1, sha1);
1475                spec->sha1_valid = !is_null_sha1(sha1);
1476        }
1477}
1478
1479/*
1480 * Given a name and sha1 pair, if the index tells us the file in
1481 * the work tree has that object contents, return true, so that
1482 * prepare_temp_file() does not have to inflate and extract.
1483 */
1484static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1485{
1486        struct cache_entry *ce;
1487        struct stat st;
1488        int pos, len;
1489
1490        /* We do not read the cache ourselves here, because the
1491         * benchmark with my previous version that always reads cache
1492         * shows that it makes things worse for diff-tree comparing
1493         * two linux-2.6 kernel trees in an already checked out work
1494         * tree.  This is because most diff-tree comparisons deal with
1495         * only a small number of files, while reading the cache is
1496         * expensive for a large project, and its cost outweighs the
1497         * savings we get by not inflating the object to a temporary
1498         * file.  Practically, this code only helps when we are used
1499         * by diff-cache --cached, which does read the cache before
1500         * calling us.
1501         */
1502        if (!active_cache)
1503                return 0;
1504
1505        /* We want to avoid the working directory if our caller
1506         * doesn't need the data in a normal file, this system
1507         * is rather slow with its stat/open/mmap/close syscalls,
1508         * and the object is contained in a pack file.  The pack
1509         * is probably already open and will be faster to obtain
1510         * the data through than the working directory.  Loose
1511         * objects however would tend to be slower as they need
1512         * to be individually opened and inflated.
1513         */
1514        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1515                return 0;
1516
1517        len = strlen(name);
1518        pos = cache_name_pos(name, len);
1519        if (pos < 0)
1520                return 0;
1521        ce = active_cache[pos];
1522        if ((lstat(name, &st) < 0) ||
1523            !S_ISREG(st.st_mode) || /* careful! */
1524            ce_match_stat(ce, &st, 0) ||
1525            hashcmp(sha1, ce->sha1))
1526                return 0;
1527        /* we return 1 only when we can stat, it is a regular file,
1528         * stat information matches, and sha1 recorded in the cache
1529         * matches.  I.e. we know the file in the work tree really is
1530         * the same as the <name, sha1> pair.
1531         */
1532        return 1;
1533}
1534
1535static int populate_from_stdin(struct diff_filespec *s)
1536{
1537        struct strbuf buf;
1538        size_t size = 0;
1539
1540        strbuf_init(&buf, 0);
1541        if (strbuf_read(&buf, 0, 0) < 0)
1542                return error("error while reading from stdin %s",
1543                                     strerror(errno));
1544
1545        s->should_munmap = 0;
1546        s->data = strbuf_detach(&buf, &size);
1547        s->size = size;
1548        s->should_free = 1;
1549        return 0;
1550}
1551
1552static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1553{
1554        int len;
1555        char *data = xmalloc(100);
1556        len = snprintf(data, 100,
1557                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1558        s->data = data;
1559        s->size = len;
1560        s->should_free = 1;
1561        if (size_only) {
1562                s->data = NULL;
1563                free(data);
1564        }
1565        return 0;
1566}
1567
1568/*
1569 * While doing rename detection and pickaxe operation, we may need to
1570 * grab the data for the blob (or file) for our own in-core comparison.
1571 * diff_filespec has data and size fields for this purpose.
1572 */
1573int diff_populate_filespec(struct diff_filespec *s, int size_only)
1574{
1575        int err = 0;
1576        if (!DIFF_FILE_VALID(s))
1577                die("internal error: asking to populate invalid file.");
1578        if (S_ISDIR(s->mode))
1579                return -1;
1580
1581        if (s->data)
1582                return 0;
1583
1584        if (size_only && 0 < s->size)
1585                return 0;
1586
1587        if (S_ISGITLINK(s->mode))
1588                return diff_populate_gitlink(s, size_only);
1589
1590        if (!s->sha1_valid ||
1591            reuse_worktree_file(s->path, s->sha1, 0)) {
1592                struct strbuf buf;
1593                struct stat st;
1594                int fd;
1595
1596                if (!strcmp(s->path, "-"))
1597                        return populate_from_stdin(s);
1598
1599                if (lstat(s->path, &st) < 0) {
1600                        if (errno == ENOENT) {
1601                        err_empty:
1602                                err = -1;
1603                        empty:
1604                                s->data = (char *)"";
1605                                s->size = 0;
1606                                return err;
1607                        }
1608                }
1609                s->size = xsize_t(st.st_size);
1610                if (!s->size)
1611                        goto empty;
1612                if (size_only)
1613                        return 0;
1614                if (S_ISLNK(st.st_mode)) {
1615                        int ret;
1616                        s->data = xmalloc(s->size);
1617                        s->should_free = 1;
1618                        ret = readlink(s->path, s->data, s->size);
1619                        if (ret < 0) {
1620                                free(s->data);
1621                                goto err_empty;
1622                        }
1623                        return 0;
1624                }
1625                fd = open(s->path, O_RDONLY);
1626                if (fd < 0)
1627                        goto err_empty;
1628                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1629                close(fd);
1630                s->should_munmap = 1;
1631
1632                /*
1633                 * Convert from working tree format to canonical git format
1634                 */
1635                strbuf_init(&buf, 0);
1636                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1637                        size_t size = 0;
1638                        munmap(s->data, s->size);
1639                        s->should_munmap = 0;
1640                        s->data = strbuf_detach(&buf, &size);
1641                        s->size = size;
1642                        s->should_free = 1;
1643                }
1644        }
1645        else {
1646                enum object_type type;
1647                if (size_only)
1648                        type = sha1_object_info(s->sha1, &s->size);
1649                else {
1650                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1651                        s->should_free = 1;
1652                }
1653        }
1654        return 0;
1655}
1656
1657void diff_free_filespec_blob(struct diff_filespec *s)
1658{
1659        if (s->should_free)
1660                free(s->data);
1661        else if (s->should_munmap)
1662                munmap(s->data, s->size);
1663
1664        if (s->should_free || s->should_munmap) {
1665                s->should_free = s->should_munmap = 0;
1666                s->data = NULL;
1667        }
1668}
1669
1670void diff_free_filespec_data(struct diff_filespec *s)
1671{
1672        diff_free_filespec_blob(s);
1673        free(s->cnt_data);
1674        s->cnt_data = NULL;
1675}
1676
1677static void prep_temp_blob(struct diff_tempfile *temp,
1678                           void *blob,
1679                           unsigned long size,
1680                           const unsigned char *sha1,
1681                           int mode)
1682{
1683        int fd;
1684
1685        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1686        if (fd < 0)
1687                die("unable to create temp-file: %s", strerror(errno));
1688        if (write_in_full(fd, blob, size) != size)
1689                die("unable to write temp-file");
1690        close(fd);
1691        temp->name = temp->tmp_path;
1692        strcpy(temp->hex, sha1_to_hex(sha1));
1693        temp->hex[40] = 0;
1694        sprintf(temp->mode, "%06o", mode);
1695}
1696
1697static void prepare_temp_file(const char *name,
1698                              struct diff_tempfile *temp,
1699                              struct diff_filespec *one)
1700{
1701        if (!DIFF_FILE_VALID(one)) {
1702        not_a_valid_file:
1703                /* A '-' entry produces this for file-2, and
1704                 * a '+' entry produces this for file-1.
1705                 */
1706                temp->name = "/dev/null";
1707                strcpy(temp->hex, ".");
1708                strcpy(temp->mode, ".");
1709                return;
1710        }
1711
1712        if (!one->sha1_valid ||
1713            reuse_worktree_file(name, one->sha1, 1)) {
1714                struct stat st;
1715                if (lstat(name, &st) < 0) {
1716                        if (errno == ENOENT)
1717                                goto not_a_valid_file;
1718                        die("stat(%s): %s", name, strerror(errno));
1719                }
1720                if (S_ISLNK(st.st_mode)) {
1721                        int ret;
1722                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1723                        size_t sz = xsize_t(st.st_size);
1724                        if (sizeof(buf) <= st.st_size)
1725                                die("symlink too long: %s", name);
1726                        ret = readlink(name, buf, sz);
1727                        if (ret < 0)
1728                                die("readlink(%s)", name);
1729                        prep_temp_blob(temp, buf, sz,
1730                                       (one->sha1_valid ?
1731                                        one->sha1 : null_sha1),
1732                                       (one->sha1_valid ?
1733                                        one->mode : S_IFLNK));
1734                }
1735                else {
1736                        /* we can borrow from the file in the work tree */
1737                        temp->name = name;
1738                        if (!one->sha1_valid)
1739                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1740                        else
1741                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1742                        /* Even though we may sometimes borrow the
1743                         * contents from the work tree, we always want
1744                         * one->mode.  mode is trustworthy even when
1745                         * !(one->sha1_valid), as long as
1746                         * DIFF_FILE_VALID(one).
1747                         */
1748                        sprintf(temp->mode, "%06o", one->mode);
1749                }
1750                return;
1751        }
1752        else {
1753                if (diff_populate_filespec(one, 0))
1754                        die("cannot read data blob for %s", one->path);
1755                prep_temp_blob(temp, one->data, one->size,
1756                               one->sha1, one->mode);
1757        }
1758}
1759
1760static void remove_tempfile(void)
1761{
1762        int i;
1763
1764        for (i = 0; i < 2; i++)
1765                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1766                        unlink(diff_temp[i].name);
1767                        diff_temp[i].name = NULL;
1768                }
1769}
1770
1771static void remove_tempfile_on_signal(int signo)
1772{
1773        remove_tempfile();
1774        signal(SIGINT, SIG_DFL);
1775        raise(signo);
1776}
1777
1778/* An external diff command takes:
1779 *
1780 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1781 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1782 *
1783 */
1784static void run_external_diff(const char *pgm,
1785                              const char *name,
1786                              const char *other,
1787                              struct diff_filespec *one,
1788                              struct diff_filespec *two,
1789                              const char *xfrm_msg,
1790                              int complete_rewrite)
1791{
1792        const char *spawn_arg[10];
1793        struct diff_tempfile *temp = diff_temp;
1794        int retval;
1795        static int atexit_asked = 0;
1796        const char *othername;
1797        const char **arg = &spawn_arg[0];
1798
1799        othername = (other? other : name);
1800        if (one && two) {
1801                prepare_temp_file(name, &temp[0], one);
1802                prepare_temp_file(othername, &temp[1], two);
1803                if (! atexit_asked &&
1804                    (temp[0].name == temp[0].tmp_path ||
1805                     temp[1].name == temp[1].tmp_path)) {
1806                        atexit_asked = 1;
1807                        atexit(remove_tempfile);
1808                }
1809                signal(SIGINT, remove_tempfile_on_signal);
1810        }
1811
1812        if (one && two) {
1813                *arg++ = pgm;
1814                *arg++ = name;
1815                *arg++ = temp[0].name;
1816                *arg++ = temp[0].hex;
1817                *arg++ = temp[0].mode;
1818                *arg++ = temp[1].name;
1819                *arg++ = temp[1].hex;
1820                *arg++ = temp[1].mode;
1821                if (other) {
1822                        *arg++ = other;
1823                        *arg++ = xfrm_msg;
1824                }
1825        } else {
1826                *arg++ = pgm;
1827                *arg++ = name;
1828        }
1829        *arg = NULL;
1830        fflush(NULL);
1831        retval = run_command_v_opt(spawn_arg, 0);
1832        remove_tempfile();
1833        if (retval) {
1834                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1835                exit(1);
1836        }
1837}
1838
1839static const char *external_diff_attr(const char *name)
1840{
1841        struct git_attr_check attr_diff_check;
1842
1843        setup_diff_attr_check(&attr_diff_check);
1844        if (!git_checkattr(name, 1, &attr_diff_check)) {
1845                const char *value = attr_diff_check.value;
1846                if (!ATTR_TRUE(value) &&
1847                    !ATTR_FALSE(value) &&
1848                    !ATTR_UNSET(value)) {
1849                        struct ll_diff_driver *drv;
1850
1851                        read_config_if_needed();
1852                        for (drv = user_diff; drv; drv = drv->next)
1853                                if (!strcmp(drv->name, value))
1854                                        return drv->cmd;
1855                }
1856        }
1857        return NULL;
1858}
1859
1860static void run_diff_cmd(const char *pgm,
1861                         const char *name,
1862                         const char *other,
1863                         struct diff_filespec *one,
1864                         struct diff_filespec *two,
1865                         const char *xfrm_msg,
1866                         struct diff_options *o,
1867                         int complete_rewrite)
1868{
1869        if (!o->allow_external)
1870                pgm = NULL;
1871        else {
1872                const char *cmd = external_diff_attr(name);
1873                if (cmd)
1874                        pgm = cmd;
1875        }
1876
1877        if (pgm) {
1878                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1879                                  complete_rewrite);
1880                return;
1881        }
1882        if (one && two)
1883                builtin_diff(name, other ? other : name,
1884                             one, two, xfrm_msg, o, complete_rewrite);
1885        else
1886                printf("* Unmerged path %s\n", name);
1887}
1888
1889static void diff_fill_sha1_info(struct diff_filespec *one)
1890{
1891        if (DIFF_FILE_VALID(one)) {
1892                if (!one->sha1_valid) {
1893                        struct stat st;
1894                        if (!strcmp(one->path, "-")) {
1895                                hashcpy(one->sha1, null_sha1);
1896                                return;
1897                        }
1898                        if (lstat(one->path, &st) < 0)
1899                                die("stat %s", one->path);
1900                        if (index_path(one->sha1, one->path, &st, 0))
1901                                die("cannot hash %s\n", one->path);
1902                }
1903        }
1904        else
1905                hashclr(one->sha1);
1906}
1907
1908static int similarity_index(struct diff_filepair *p)
1909{
1910        return p->score * 100 / MAX_SCORE;
1911}
1912
1913static void run_diff(struct diff_filepair *p, struct diff_options *o)
1914{
1915        const char *pgm = external_diff();
1916        struct strbuf msg;
1917        char *xfrm_msg;
1918        struct diff_filespec *one = p->one;
1919        struct diff_filespec *two = p->two;
1920        const char *name;
1921        const char *other;
1922        int complete_rewrite = 0;
1923
1924
1925        if (DIFF_PAIR_UNMERGED(p)) {
1926                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1927                return;
1928        }
1929
1930        name  = p->one->path;
1931        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1932        diff_fill_sha1_info(one);
1933        diff_fill_sha1_info(two);
1934
1935        strbuf_init(&msg, PATH_MAX * 2 + 300);
1936        switch (p->status) {
1937        case DIFF_STATUS_COPIED:
1938                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1939                strbuf_addstr(&msg, "\ncopy from ");
1940                quote_c_style(name, &msg, NULL, 0);
1941                strbuf_addstr(&msg, "\ncopy to ");
1942                quote_c_style(other, &msg, NULL, 0);
1943                strbuf_addch(&msg, '\n');
1944                break;
1945        case DIFF_STATUS_RENAMED:
1946                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1947                strbuf_addstr(&msg, "\nrename from ");
1948                quote_c_style(name, &msg, NULL, 0);
1949                strbuf_addstr(&msg, "\nrename to ");
1950                quote_c_style(other, &msg, NULL, 0);
1951                strbuf_addch(&msg, '\n');
1952                break;
1953        case DIFF_STATUS_MODIFIED:
1954                if (p->score) {
1955                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
1956                                        similarity_index(p));
1957                        complete_rewrite = 1;
1958                        break;
1959                }
1960                /* fallthru */
1961        default:
1962                /* nothing */
1963                ;
1964        }
1965
1966        if (hashcmp(one->sha1, two->sha1)) {
1967                int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
1968
1969                if (o->binary) {
1970                        mmfile_t mf;
1971                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
1972                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
1973                                abbrev = 40;
1974                }
1975                strbuf_addf(&msg, "index %.*s..%.*s",
1976                                abbrev, sha1_to_hex(one->sha1),
1977                                abbrev, sha1_to_hex(two->sha1));
1978                if (one->mode == two->mode)
1979                        strbuf_addf(&msg, " %06o", one->mode);
1980                strbuf_addch(&msg, '\n');
1981        }
1982
1983        if (msg.len)
1984                strbuf_setlen(&msg, msg.len - 1);
1985        xfrm_msg = msg.len ? msg.buf : NULL;
1986
1987        if (!pgm &&
1988            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1989            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1990                /* a filepair that changes between file and symlink
1991                 * needs to be split into deletion and creation.
1992                 */
1993                struct diff_filespec *null = alloc_filespec(two->path);
1994                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1995                free(null);
1996                null = alloc_filespec(one->path);
1997                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1998                free(null);
1999        }
2000        else
2001                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
2002                             complete_rewrite);
2003
2004        strbuf_release(&msg);
2005}
2006
2007static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2008                         struct diffstat_t *diffstat)
2009{
2010        const char *name;
2011        const char *other;
2012        int complete_rewrite = 0;
2013
2014        if (DIFF_PAIR_UNMERGED(p)) {
2015                /* unmerged */
2016                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2017                return;
2018        }
2019
2020        name = p->one->path;
2021        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2022
2023        diff_fill_sha1_info(p->one);
2024        diff_fill_sha1_info(p->two);
2025
2026        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2027                complete_rewrite = 1;
2028        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2029}
2030
2031static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2032{
2033        const char *name;
2034        const char *other;
2035
2036        if (DIFF_PAIR_UNMERGED(p)) {
2037                /* unmerged */
2038                return;
2039        }
2040
2041        name = p->one->path;
2042        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2043
2044        diff_fill_sha1_info(p->one);
2045        diff_fill_sha1_info(p->two);
2046
2047        builtin_checkdiff(name, other, p->one, p->two, o);
2048}
2049
2050void diff_setup(struct diff_options *options)
2051{
2052        memset(options, 0, sizeof(*options));
2053        options->line_termination = '\n';
2054        options->break_opt = -1;
2055        options->rename_limit = -1;
2056        options->context = 3;
2057        options->msg_sep = "";
2058
2059        options->change = diff_change;
2060        options->add_remove = diff_addremove;
2061        options->color_diff = diff_use_color_default;
2062        options->detect_rename = diff_detect_rename_default;
2063}
2064
2065int diff_setup_done(struct diff_options *options)
2066{
2067        int count = 0;
2068
2069        if (options->output_format & DIFF_FORMAT_NAME)
2070                count++;
2071        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2072                count++;
2073        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2074                count++;
2075        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2076                count++;
2077        if (count > 1)
2078                die("--name-only, --name-status, --check and -s are mutually exclusive");
2079
2080        if (options->find_copies_harder)
2081                options->detect_rename = DIFF_DETECT_COPY;
2082
2083        if (options->output_format & (DIFF_FORMAT_NAME |
2084                                      DIFF_FORMAT_NAME_STATUS |
2085                                      DIFF_FORMAT_CHECKDIFF |
2086                                      DIFF_FORMAT_NO_OUTPUT))
2087                options->output_format &= ~(DIFF_FORMAT_RAW |
2088                                            DIFF_FORMAT_NUMSTAT |
2089                                            DIFF_FORMAT_DIFFSTAT |
2090                                            DIFF_FORMAT_SHORTSTAT |
2091                                            DIFF_FORMAT_SUMMARY |
2092                                            DIFF_FORMAT_PATCH);
2093
2094        /*
2095         * These cases always need recursive; we do not drop caller-supplied
2096         * recursive bits for other formats here.
2097         */
2098        if (options->output_format & (DIFF_FORMAT_PATCH |
2099                                      DIFF_FORMAT_NUMSTAT |
2100                                      DIFF_FORMAT_DIFFSTAT |
2101                                      DIFF_FORMAT_SHORTSTAT |
2102                                      DIFF_FORMAT_SUMMARY |
2103                                      DIFF_FORMAT_CHECKDIFF))
2104                options->recursive = 1;
2105        /*
2106         * Also pickaxe would not work very well if you do not say recursive
2107         */
2108        if (options->pickaxe)
2109                options->recursive = 1;
2110
2111        if (options->detect_rename && options->rename_limit < 0)
2112                options->rename_limit = diff_rename_limit_default;
2113        if (options->setup & DIFF_SETUP_USE_CACHE) {
2114                if (!active_cache)
2115                        /* read-cache does not die even when it fails
2116                         * so it is safe for us to do this here.  Also
2117                         * it does not smudge active_cache or active_nr
2118                         * when it fails, so we do not have to worry about
2119                         * cleaning it up ourselves either.
2120                         */
2121                        read_cache();
2122        }
2123        if (options->abbrev <= 0 || 40 < options->abbrev)
2124                options->abbrev = 40; /* full */
2125
2126        /*
2127         * It does not make sense to show the first hit we happened
2128         * to have found.  It does not make sense not to return with
2129         * exit code in such a case either.
2130         */
2131        if (options->quiet) {
2132                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2133                options->exit_with_status = 1;
2134        }
2135
2136        /*
2137         * If we postprocess in diffcore, we cannot simply return
2138         * upon the first hit.  We need to run diff as usual.
2139         */
2140        if (options->pickaxe || options->filter)
2141                options->quiet = 0;
2142
2143        return 0;
2144}
2145
2146static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2147{
2148        char c, *eq;
2149        int len;
2150
2151        if (*arg != '-')
2152                return 0;
2153        c = *++arg;
2154        if (!c)
2155                return 0;
2156        if (c == arg_short) {
2157                c = *++arg;
2158                if (!c)
2159                        return 1;
2160                if (val && isdigit(c)) {
2161                        char *end;
2162                        int n = strtoul(arg, &end, 10);
2163                        if (*end)
2164                                return 0;
2165                        *val = n;
2166                        return 1;
2167                }
2168                return 0;
2169        }
2170        if (c != '-')
2171                return 0;
2172        arg++;
2173        eq = strchr(arg, '=');
2174        if (eq)
2175                len = eq - arg;
2176        else
2177                len = strlen(arg);
2178        if (!len || strncmp(arg, arg_long, len))
2179                return 0;
2180        if (eq) {
2181                int n;
2182                char *end;
2183                if (!isdigit(*++eq))
2184                        return 0;
2185                n = strtoul(eq, &end, 10);
2186                if (*end)
2187                        return 0;
2188                *val = n;
2189        }
2190        return 1;
2191}
2192
2193static int diff_scoreopt_parse(const char *opt);
2194
2195int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2196{
2197        const char *arg = av[0];
2198        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2199                options->output_format |= DIFF_FORMAT_PATCH;
2200        else if (opt_arg(arg, 'U', "unified", &options->context))
2201                options->output_format |= DIFF_FORMAT_PATCH;
2202        else if (!strcmp(arg, "--raw"))
2203                options->output_format |= DIFF_FORMAT_RAW;
2204        else if (!strcmp(arg, "--patch-with-raw")) {
2205                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2206        }
2207        else if (!strcmp(arg, "--numstat")) {
2208                options->output_format |= DIFF_FORMAT_NUMSTAT;
2209        }
2210        else if (!strcmp(arg, "--shortstat")) {
2211                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2212        }
2213        else if (!prefixcmp(arg, "--stat")) {
2214                char *end;
2215                int width = options->stat_width;
2216                int name_width = options->stat_name_width;
2217                arg += 6;
2218                end = (char *)arg;
2219
2220                switch (*arg) {
2221                case '-':
2222                        if (!prefixcmp(arg, "-width="))
2223                                width = strtoul(arg + 7, &end, 10);
2224                        else if (!prefixcmp(arg, "-name-width="))
2225                                name_width = strtoul(arg + 12, &end, 10);
2226                        break;
2227                case '=':
2228                        width = strtoul(arg+1, &end, 10);
2229                        if (*end == ',')
2230                                name_width = strtoul(end+1, &end, 10);
2231                }
2232
2233                /* Important! This checks all the error cases! */
2234                if (*end)
2235                        return 0;
2236                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2237                options->stat_name_width = name_width;
2238                options->stat_width = width;
2239        }
2240        else if (!strcmp(arg, "--check"))
2241                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2242        else if (!strcmp(arg, "--summary"))
2243                options->output_format |= DIFF_FORMAT_SUMMARY;
2244        else if (!strcmp(arg, "--patch-with-stat")) {
2245                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2246        }
2247        else if (!strcmp(arg, "-z"))
2248                options->line_termination = 0;
2249        else if (!prefixcmp(arg, "-l"))
2250                options->rename_limit = strtoul(arg+2, NULL, 10);
2251        else if (!strcmp(arg, "--full-index"))
2252                options->full_index = 1;
2253        else if (!strcmp(arg, "--binary")) {
2254                options->output_format |= DIFF_FORMAT_PATCH;
2255                options->binary = 1;
2256        }
2257        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text")) {
2258                options->text = 1;
2259        }
2260        else if (!strcmp(arg, "--name-only"))
2261                options->output_format |= DIFF_FORMAT_NAME;
2262        else if (!strcmp(arg, "--name-status"))
2263                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2264        else if (!strcmp(arg, "-R"))
2265                options->reverse_diff = 1;
2266        else if (!prefixcmp(arg, "-S"))
2267                options->pickaxe = arg + 2;
2268        else if (!strcmp(arg, "-s")) {
2269                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2270        }
2271        else if (!prefixcmp(arg, "-O"))
2272                options->orderfile = arg + 2;
2273        else if (!prefixcmp(arg, "--diff-filter="))
2274                options->filter = arg + 14;
2275        else if (!strcmp(arg, "--pickaxe-all"))
2276                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2277        else if (!strcmp(arg, "--pickaxe-regex"))
2278                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2279        else if (!prefixcmp(arg, "-B")) {
2280                if ((options->break_opt =
2281                     diff_scoreopt_parse(arg)) == -1)
2282                        return -1;
2283        }
2284        else if (!prefixcmp(arg, "-M")) {
2285                if ((options->rename_score =
2286                     diff_scoreopt_parse(arg)) == -1)
2287                        return -1;
2288                options->detect_rename = DIFF_DETECT_RENAME;
2289        }
2290        else if (!prefixcmp(arg, "-C")) {
2291                if (options->detect_rename == DIFF_DETECT_COPY)
2292                        options->find_copies_harder = 1;
2293                if ((options->rename_score =
2294                     diff_scoreopt_parse(arg)) == -1)
2295                        return -1;
2296                options->detect_rename = DIFF_DETECT_COPY;
2297        }
2298        else if (!strcmp(arg, "--find-copies-harder"))
2299                options->find_copies_harder = 1;
2300        else if (!strcmp(arg, "--follow"))
2301                options->follow_renames = 1;
2302        else if (!strcmp(arg, "--abbrev"))
2303                options->abbrev = DEFAULT_ABBREV;
2304        else if (!prefixcmp(arg, "--abbrev=")) {
2305                options->abbrev = strtoul(arg + 9, NULL, 10);
2306                if (options->abbrev < MINIMUM_ABBREV)
2307                        options->abbrev = MINIMUM_ABBREV;
2308                else if (40 < options->abbrev)
2309                        options->abbrev = 40;
2310        }
2311        else if (!strcmp(arg, "--color"))
2312                options->color_diff = 1;
2313        else if (!strcmp(arg, "--no-color"))
2314                options->color_diff = 0;
2315        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2316                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2317        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2318                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2319        else if (!strcmp(arg, "--ignore-space-at-eol"))
2320                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2321        else if (!strcmp(arg, "--color-words"))
2322                options->color_diff = options->color_diff_words = 1;
2323        else if (!strcmp(arg, "--no-renames"))
2324                options->detect_rename = 0;
2325        else if (!strcmp(arg, "--exit-code"))
2326                options->exit_with_status = 1;
2327        else if (!strcmp(arg, "--quiet"))
2328                options->quiet = 1;
2329        else if (!strcmp(arg, "--ext-diff"))
2330                options->allow_external = 1;
2331        else if (!strcmp(arg, "--no-ext-diff"))
2332                options->allow_external = 0;
2333        else
2334                return 0;
2335        return 1;
2336}
2337
2338static int parse_num(const char **cp_p)
2339{
2340        unsigned long num, scale;
2341        int ch, dot;
2342        const char *cp = *cp_p;
2343
2344        num = 0;
2345        scale = 1;
2346        dot = 0;
2347        for(;;) {
2348                ch = *cp;
2349                if ( !dot && ch == '.' ) {
2350                        scale = 1;
2351                        dot = 1;
2352                } else if ( ch == '%' ) {
2353                        scale = dot ? scale*100 : 100;
2354                        cp++;   /* % is always at the end */
2355                        break;
2356                } else if ( ch >= '0' && ch <= '9' ) {
2357                        if ( scale < 100000 ) {
2358                                scale *= 10;
2359                                num = (num*10) + (ch-'0');
2360                        }
2361                } else {
2362                        break;
2363                }
2364                cp++;
2365        }
2366        *cp_p = cp;
2367
2368        /* user says num divided by scale and we say internally that
2369         * is MAX_SCORE * num / scale.
2370         */
2371        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2372}
2373
2374static int diff_scoreopt_parse(const char *opt)
2375{
2376        int opt1, opt2, cmd;
2377
2378        if (*opt++ != '-')
2379                return -1;
2380        cmd = *opt++;
2381        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2382                return -1; /* that is not a -M, -C nor -B option */
2383
2384        opt1 = parse_num(&opt);
2385        if (cmd != 'B')
2386                opt2 = 0;
2387        else {
2388                if (*opt == 0)
2389                        opt2 = 0;
2390                else if (*opt != '/')
2391                        return -1; /* we expect -B80/99 or -B80 */
2392                else {
2393                        opt++;
2394                        opt2 = parse_num(&opt);
2395                }
2396        }
2397        if (*opt != 0)
2398                return -1;
2399        return opt1 | (opt2 << 16);
2400}
2401
2402struct diff_queue_struct diff_queued_diff;
2403
2404void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2405{
2406        if (queue->alloc <= queue->nr) {
2407                queue->alloc = alloc_nr(queue->alloc);
2408                queue->queue = xrealloc(queue->queue,
2409                                        sizeof(dp) * queue->alloc);
2410        }
2411        queue->queue[queue->nr++] = dp;
2412}
2413
2414struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2415                                 struct diff_filespec *one,
2416                                 struct diff_filespec *two)
2417{
2418        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2419        dp->one = one;
2420        dp->two = two;
2421        if (queue)
2422                diff_q(queue, dp);
2423        return dp;
2424}
2425
2426void diff_free_filepair(struct diff_filepair *p)
2427{
2428        free_filespec(p->one);
2429        free_filespec(p->two);
2430        free(p);
2431}
2432
2433/* This is different from find_unique_abbrev() in that
2434 * it stuffs the result with dots for alignment.
2435 */
2436const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2437{
2438        int abblen;
2439        const char *abbrev;
2440        if (len == 40)
2441                return sha1_to_hex(sha1);
2442
2443        abbrev = find_unique_abbrev(sha1, len);
2444        if (!abbrev)
2445                return sha1_to_hex(sha1);
2446        abblen = strlen(abbrev);
2447        if (abblen < 37) {
2448                static char hex[41];
2449                if (len < abblen && abblen <= len + 2)
2450                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2451                else
2452                        sprintf(hex, "%s...", abbrev);
2453                return hex;
2454        }
2455        return sha1_to_hex(sha1);
2456}
2457
2458static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2459{
2460        int line_termination = opt->line_termination;
2461        int inter_name_termination = line_termination ? '\t' : '\0';
2462
2463        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2464                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2465                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2466                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2467        }
2468        if (p->score) {
2469                printf("%c%03d%c", p->status, similarity_index(p),
2470                           inter_name_termination);
2471        } else {
2472                printf("%c%c", p->status, inter_name_termination);
2473        }
2474
2475        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2476                write_name_quoted(p->one->path, stdout, inter_name_termination);
2477                write_name_quoted(p->two->path, stdout, line_termination);
2478        } else {
2479                const char *path = p->one->mode ? p->one->path : p->two->path;
2480                write_name_quoted(path, stdout, line_termination);
2481        }
2482}
2483
2484int diff_unmodified_pair(struct diff_filepair *p)
2485{
2486        /* This function is written stricter than necessary to support
2487         * the currently implemented transformers, but the idea is to
2488         * let transformers to produce diff_filepairs any way they want,
2489         * and filter and clean them up here before producing the output.
2490         */
2491        struct diff_filespec *one = p->one, *two = p->two;
2492
2493        if (DIFF_PAIR_UNMERGED(p))
2494                return 0; /* unmerged is interesting */
2495
2496        /* deletion, addition, mode or type change
2497         * and rename are all interesting.
2498         */
2499        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2500            DIFF_PAIR_MODE_CHANGED(p) ||
2501            strcmp(one->path, two->path))
2502                return 0;
2503
2504        /* both are valid and point at the same path.  that is, we are
2505         * dealing with a change.
2506         */
2507        if (one->sha1_valid && two->sha1_valid &&
2508            !hashcmp(one->sha1, two->sha1))
2509                return 1; /* no change */
2510        if (!one->sha1_valid && !two->sha1_valid)
2511                return 1; /* both look at the same file on the filesystem. */
2512        return 0;
2513}
2514
2515static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2516{
2517        if (diff_unmodified_pair(p))
2518                return;
2519
2520        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2521            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2522                return; /* no tree diffs in patch format */
2523
2524        run_diff(p, o);
2525}
2526
2527static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2528                            struct diffstat_t *diffstat)
2529{
2530        if (diff_unmodified_pair(p))
2531                return;
2532
2533        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2534            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2535                return; /* no tree diffs in patch format */
2536
2537        run_diffstat(p, o, diffstat);
2538}
2539
2540static void diff_flush_checkdiff(struct diff_filepair *p,
2541                struct diff_options *o)
2542{
2543        if (diff_unmodified_pair(p))
2544                return;
2545
2546        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2547            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2548                return; /* no tree diffs in patch format */
2549
2550        run_checkdiff(p, o);
2551}
2552
2553int diff_queue_is_empty(void)
2554{
2555        struct diff_queue_struct *q = &diff_queued_diff;
2556        int i;
2557        for (i = 0; i < q->nr; i++)
2558                if (!diff_unmodified_pair(q->queue[i]))
2559                        return 0;
2560        return 1;
2561}
2562
2563#if DIFF_DEBUG
2564void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2565{
2566        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2567                x, one ? one : "",
2568                s->path,
2569                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2570                s->mode,
2571                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2572        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2573                x, one ? one : "",
2574                s->size, s->xfrm_flags);
2575}
2576
2577void diff_debug_filepair(const struct diff_filepair *p, int i)
2578{
2579        diff_debug_filespec(p->one, i, "one");
2580        diff_debug_filespec(p->two, i, "two");
2581        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2582                p->score, p->status ? p->status : '?',
2583                p->one->rename_used, p->broken_pair);
2584}
2585
2586void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2587{
2588        int i;
2589        if (msg)
2590                fprintf(stderr, "%s\n", msg);
2591        fprintf(stderr, "q->nr = %d\n", q->nr);
2592        for (i = 0; i < q->nr; i++) {
2593                struct diff_filepair *p = q->queue[i];
2594                diff_debug_filepair(p, i);
2595        }
2596}
2597#endif
2598
2599static void diff_resolve_rename_copy(void)
2600{
2601        int i;
2602        struct diff_filepair *p;
2603        struct diff_queue_struct *q = &diff_queued_diff;
2604
2605        diff_debug_queue("resolve-rename-copy", q);
2606
2607        for (i = 0; i < q->nr; i++) {
2608                p = q->queue[i];
2609                p->status = 0; /* undecided */
2610                if (DIFF_PAIR_UNMERGED(p))
2611                        p->status = DIFF_STATUS_UNMERGED;
2612                else if (!DIFF_FILE_VALID(p->one))
2613                        p->status = DIFF_STATUS_ADDED;
2614                else if (!DIFF_FILE_VALID(p->two))
2615                        p->status = DIFF_STATUS_DELETED;
2616                else if (DIFF_PAIR_TYPE_CHANGED(p))
2617                        p->status = DIFF_STATUS_TYPE_CHANGED;
2618
2619                /* from this point on, we are dealing with a pair
2620                 * whose both sides are valid and of the same type, i.e.
2621                 * either in-place edit or rename/copy edit.
2622                 */
2623                else if (DIFF_PAIR_RENAME(p)) {
2624                        /*
2625                         * A rename might have re-connected a broken
2626                         * pair up, causing the pathnames to be the
2627                         * same again. If so, that's not a rename at
2628                         * all, just a modification..
2629                         *
2630                         * Otherwise, see if this source was used for
2631                         * multiple renames, in which case we decrement
2632                         * the count, and call it a copy.
2633                         */
2634                        if (!strcmp(p->one->path, p->two->path))
2635                                p->status = DIFF_STATUS_MODIFIED;
2636                        else if (--p->one->rename_used > 0)
2637                                p->status = DIFF_STATUS_COPIED;
2638                        else
2639                                p->status = DIFF_STATUS_RENAMED;
2640                }
2641                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2642                         p->one->mode != p->two->mode ||
2643                         is_null_sha1(p->one->sha1))
2644                        p->status = DIFF_STATUS_MODIFIED;
2645                else {
2646                        /* This is a "no-change" entry and should not
2647                         * happen anymore, but prepare for broken callers.
2648                         */
2649                        error("feeding unmodified %s to diffcore",
2650                              p->one->path);
2651                        p->status = DIFF_STATUS_UNKNOWN;
2652                }
2653        }
2654        diff_debug_queue("resolve-rename-copy done", q);
2655}
2656
2657static int check_pair_status(struct diff_filepair *p)
2658{
2659        switch (p->status) {
2660        case DIFF_STATUS_UNKNOWN:
2661                return 0;
2662        case 0:
2663                die("internal error in diff-resolve-rename-copy");
2664        default:
2665                return 1;
2666        }
2667}
2668
2669static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2670{
2671        int fmt = opt->output_format;
2672
2673        if (fmt & DIFF_FORMAT_CHECKDIFF)
2674                diff_flush_checkdiff(p, opt);
2675        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2676                diff_flush_raw(p, opt);
2677        else if (fmt & DIFF_FORMAT_NAME)
2678                write_name_quoted(p->two->path, stdout, opt->line_termination);
2679}
2680
2681static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2682{
2683        if (fs->mode)
2684                printf(" %s mode %06o ", newdelete, fs->mode);
2685        else
2686                printf(" %s ", newdelete);
2687        write_name_quoted(fs->path, stdout, '\n');
2688}
2689
2690
2691static void show_mode_change(struct diff_filepair *p, int show_name)
2692{
2693        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2694                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2695                        show_name ? ' ' : '\n');
2696                if (show_name) {
2697                        write_name_quoted(p->two->path, stdout, '\n');
2698                }
2699        }
2700}
2701
2702static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2703{
2704        char *names = pprint_rename(p->one->path, p->two->path);
2705
2706        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2707        free(names);
2708        show_mode_change(p, 0);
2709}
2710
2711static void diff_summary(struct diff_filepair *p)
2712{
2713        switch(p->status) {
2714        case DIFF_STATUS_DELETED:
2715                show_file_mode_name("delete", p->one);
2716                break;
2717        case DIFF_STATUS_ADDED:
2718                show_file_mode_name("create", p->two);
2719                break;
2720        case DIFF_STATUS_COPIED:
2721                show_rename_copy("copy", p);
2722                break;
2723        case DIFF_STATUS_RENAMED:
2724                show_rename_copy("rename", p);
2725                break;
2726        default:
2727                if (p->score) {
2728                        puts(" rewrite ");
2729                        write_name_quoted(p->two->path, stdout, ' ');
2730                        printf("(%d%%)\n", similarity_index(p));
2731                }
2732                show_mode_change(p, !p->score);
2733                break;
2734        }
2735}
2736
2737struct patch_id_t {
2738        struct xdiff_emit_state xm;
2739        SHA_CTX *ctx;
2740        int patchlen;
2741};
2742
2743static int remove_space(char *line, int len)
2744{
2745        int i;
2746        char *dst = line;
2747        unsigned char c;
2748
2749        for (i = 0; i < len; i++)
2750                if (!isspace((c = line[i])))
2751                        *dst++ = c;
2752
2753        return dst - line;
2754}
2755
2756static void patch_id_consume(void *priv, char *line, unsigned long len)
2757{
2758        struct patch_id_t *data = priv;
2759        int new_len;
2760
2761        /* Ignore line numbers when computing the SHA1 of the patch */
2762        if (!prefixcmp(line, "@@ -"))
2763                return;
2764
2765        new_len = remove_space(line, len);
2766
2767        SHA1_Update(data->ctx, line, new_len);
2768        data->patchlen += new_len;
2769}
2770
2771/* returns 0 upon success, and writes result into sha1 */
2772static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2773{
2774        struct diff_queue_struct *q = &diff_queued_diff;
2775        int i;
2776        SHA_CTX ctx;
2777        struct patch_id_t data;
2778        char buffer[PATH_MAX * 4 + 20];
2779
2780        SHA1_Init(&ctx);
2781        memset(&data, 0, sizeof(struct patch_id_t));
2782        data.ctx = &ctx;
2783        data.xm.consume = patch_id_consume;
2784
2785        for (i = 0; i < q->nr; i++) {
2786                xpparam_t xpp;
2787                xdemitconf_t xecfg;
2788                xdemitcb_t ecb;
2789                mmfile_t mf1, mf2;
2790                struct diff_filepair *p = q->queue[i];
2791                int len1, len2;
2792
2793                memset(&xecfg, 0, sizeof(xecfg));
2794                if (p->status == 0)
2795                        return error("internal diff status error");
2796                if (p->status == DIFF_STATUS_UNKNOWN)
2797                        continue;
2798                if (diff_unmodified_pair(p))
2799                        continue;
2800                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2801                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2802                        continue;
2803                if (DIFF_PAIR_UNMERGED(p))
2804                        continue;
2805
2806                diff_fill_sha1_info(p->one);
2807                diff_fill_sha1_info(p->two);
2808                if (fill_mmfile(&mf1, p->one) < 0 ||
2809                                fill_mmfile(&mf2, p->two) < 0)
2810                        return error("unable to read files to diff");
2811
2812                len1 = remove_space(p->one->path, strlen(p->one->path));
2813                len2 = remove_space(p->two->path, strlen(p->two->path));
2814                if (p->one->mode == 0)
2815                        len1 = snprintf(buffer, sizeof(buffer),
2816                                        "diff--gita/%.*sb/%.*s"
2817                                        "newfilemode%06o"
2818                                        "---/dev/null"
2819                                        "+++b/%.*s",
2820                                        len1, p->one->path,
2821                                        len2, p->two->path,
2822                                        p->two->mode,
2823                                        len2, p->two->path);
2824                else if (p->two->mode == 0)
2825                        len1 = snprintf(buffer, sizeof(buffer),
2826                                        "diff--gita/%.*sb/%.*s"
2827                                        "deletedfilemode%06o"
2828                                        "---a/%.*s"
2829                                        "+++/dev/null",
2830                                        len1, p->one->path,
2831                                        len2, p->two->path,
2832                                        p->one->mode,
2833                                        len1, p->one->path);
2834                else
2835                        len1 = snprintf(buffer, sizeof(buffer),
2836                                        "diff--gita/%.*sb/%.*s"
2837                                        "---a/%.*s"
2838                                        "+++b/%.*s",
2839                                        len1, p->one->path,
2840                                        len2, p->two->path,
2841                                        len1, p->one->path,
2842                                        len2, p->two->path);
2843                SHA1_Update(&ctx, buffer, len1);
2844
2845                xpp.flags = XDF_NEED_MINIMAL;
2846                xecfg.ctxlen = 3;
2847                xecfg.flags = XDL_EMIT_FUNCNAMES;
2848                ecb.outf = xdiff_outf;
2849                ecb.priv = &data;
2850                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2851        }
2852
2853        SHA1_Final(sha1, &ctx);
2854        return 0;
2855}
2856
2857int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2858{
2859        struct diff_queue_struct *q = &diff_queued_diff;
2860        int i;
2861        int result = diff_get_patch_id(options, sha1);
2862
2863        for (i = 0; i < q->nr; i++)
2864                diff_free_filepair(q->queue[i]);
2865
2866        free(q->queue);
2867        q->queue = NULL;
2868        q->nr = q->alloc = 0;
2869
2870        return result;
2871}
2872
2873static int is_summary_empty(const struct diff_queue_struct *q)
2874{
2875        int i;
2876
2877        for (i = 0; i < q->nr; i++) {
2878                const struct diff_filepair *p = q->queue[i];
2879
2880                switch (p->status) {
2881                case DIFF_STATUS_DELETED:
2882                case DIFF_STATUS_ADDED:
2883                case DIFF_STATUS_COPIED:
2884                case DIFF_STATUS_RENAMED:
2885                        return 0;
2886                default:
2887                        if (p->score)
2888                                return 0;
2889                        if (p->one->mode && p->two->mode &&
2890                            p->one->mode != p->two->mode)
2891                                return 0;
2892                        break;
2893                }
2894        }
2895        return 1;
2896}
2897
2898void diff_flush(struct diff_options *options)
2899{
2900        struct diff_queue_struct *q = &diff_queued_diff;
2901        int i, output_format = options->output_format;
2902        int separator = 0;
2903
2904        /*
2905         * Order: raw, stat, summary, patch
2906         * or:    name/name-status/checkdiff (other bits clear)
2907         */
2908        if (!q->nr)
2909                goto free_queue;
2910
2911        if (output_format & (DIFF_FORMAT_RAW |
2912                             DIFF_FORMAT_NAME |
2913                             DIFF_FORMAT_NAME_STATUS |
2914                             DIFF_FORMAT_CHECKDIFF)) {
2915                for (i = 0; i < q->nr; i++) {
2916                        struct diff_filepair *p = q->queue[i];
2917                        if (check_pair_status(p))
2918                                flush_one_pair(p, options);
2919                }
2920                separator++;
2921        }
2922
2923        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2924                struct diffstat_t diffstat;
2925
2926                memset(&diffstat, 0, sizeof(struct diffstat_t));
2927                diffstat.xm.consume = diffstat_consume;
2928                for (i = 0; i < q->nr; i++) {
2929                        struct diff_filepair *p = q->queue[i];
2930                        if (check_pair_status(p))
2931                                diff_flush_stat(p, options, &diffstat);
2932                }
2933                if (output_format & DIFF_FORMAT_NUMSTAT)
2934                        show_numstat(&diffstat, options);
2935                if (output_format & DIFF_FORMAT_DIFFSTAT)
2936                        show_stats(&diffstat, options);
2937                else if (output_format & DIFF_FORMAT_SHORTSTAT)
2938                        show_shortstats(&diffstat);
2939                separator++;
2940        }
2941
2942        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2943                for (i = 0; i < q->nr; i++)
2944                        diff_summary(q->queue[i]);
2945                separator++;
2946        }
2947
2948        if (output_format & DIFF_FORMAT_PATCH) {
2949                if (separator) {
2950                        if (options->stat_sep) {
2951                                /* attach patch instead of inline */
2952                                fputs(options->stat_sep, stdout);
2953                        } else {
2954                                putchar(options->line_termination);
2955                        }
2956                }
2957
2958                for (i = 0; i < q->nr; i++) {
2959                        struct diff_filepair *p = q->queue[i];
2960                        if (check_pair_status(p))
2961                                diff_flush_patch(p, options);
2962                }
2963        }
2964
2965        if (output_format & DIFF_FORMAT_CALLBACK)
2966                options->format_callback(q, options, options->format_callback_data);
2967
2968        for (i = 0; i < q->nr; i++)
2969                diff_free_filepair(q->queue[i]);
2970free_queue:
2971        free(q->queue);
2972        q->queue = NULL;
2973        q->nr = q->alloc = 0;
2974}
2975
2976static void diffcore_apply_filter(const char *filter)
2977{
2978        int i;
2979        struct diff_queue_struct *q = &diff_queued_diff;
2980        struct diff_queue_struct outq;
2981        outq.queue = NULL;
2982        outq.nr = outq.alloc = 0;
2983
2984        if (!filter)
2985                return;
2986
2987        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2988                int found;
2989                for (i = found = 0; !found && i < q->nr; i++) {
2990                        struct diff_filepair *p = q->queue[i];
2991                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2992                             ((p->score &&
2993                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2994                              (!p->score &&
2995                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2996                            ((p->status != DIFF_STATUS_MODIFIED) &&
2997                             strchr(filter, p->status)))
2998                                found++;
2999                }
3000                if (found)
3001                        return;
3002
3003                /* otherwise we will clear the whole queue
3004                 * by copying the empty outq at the end of this
3005                 * function, but first clear the current entries
3006                 * in the queue.
3007                 */
3008                for (i = 0; i < q->nr; i++)
3009                        diff_free_filepair(q->queue[i]);
3010        }
3011        else {
3012                /* Only the matching ones */
3013                for (i = 0; i < q->nr; i++) {
3014                        struct diff_filepair *p = q->queue[i];
3015
3016                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3017                             ((p->score &&
3018                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3019                              (!p->score &&
3020                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3021                            ((p->status != DIFF_STATUS_MODIFIED) &&
3022                             strchr(filter, p->status)))
3023                                diff_q(&outq, p);
3024                        else
3025                                diff_free_filepair(p);
3026                }
3027        }
3028        free(q->queue);
3029        *q = outq;
3030}
3031
3032/* Check whether two filespecs with the same mode and size are identical */
3033static int diff_filespec_is_identical(struct diff_filespec *one,
3034                                      struct diff_filespec *two)
3035{
3036        if (S_ISGITLINK(one->mode)) {
3037                diff_fill_sha1_info(one);
3038                diff_fill_sha1_info(two);
3039                return !hashcmp(one->sha1, two->sha1);
3040        }
3041        if (diff_populate_filespec(one, 0))
3042                return 0;
3043        if (diff_populate_filespec(two, 0))
3044                return 0;
3045        return !memcmp(one->data, two->data, one->size);
3046}
3047
3048static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3049{
3050        int i;
3051        struct diff_queue_struct *q = &diff_queued_diff;
3052        struct diff_queue_struct outq;
3053        outq.queue = NULL;
3054        outq.nr = outq.alloc = 0;
3055
3056        for (i = 0; i < q->nr; i++) {
3057                struct diff_filepair *p = q->queue[i];
3058
3059                /*
3060                 * 1. Entries that come from stat info dirtyness
3061                 *    always have both sides (iow, not create/delete),
3062                 *    one side of the object name is unknown, with
3063                 *    the same mode and size.  Keep the ones that
3064                 *    do not match these criteria.  They have real
3065                 *    differences.
3066                 *
3067                 * 2. At this point, the file is known to be modified,
3068                 *    with the same mode and size, and the object
3069                 *    name of one side is unknown.  Need to inspect
3070                 *    the identical contents.
3071                 */
3072                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3073                    !DIFF_FILE_VALID(p->two) ||
3074                    (p->one->sha1_valid && p->two->sha1_valid) ||
3075                    (p->one->mode != p->two->mode) ||
3076                    diff_populate_filespec(p->one, 1) ||
3077                    diff_populate_filespec(p->two, 1) ||
3078                    (p->one->size != p->two->size) ||
3079                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3080                        diff_q(&outq, p);
3081                else {
3082                        /*
3083                         * The caller can subtract 1 from skip_stat_unmatch
3084                         * to determine how many paths were dirty only
3085                         * due to stat info mismatch.
3086                         */
3087                        if (!diffopt->no_index)
3088                                diffopt->skip_stat_unmatch++;
3089                        diff_free_filepair(p);
3090                }
3091        }
3092        free(q->queue);
3093        *q = outq;
3094}
3095
3096void diffcore_std(struct diff_options *options)
3097{
3098        if (options->quiet)
3099                return;
3100
3101        if (options->skip_stat_unmatch && !options->find_copies_harder)
3102                diffcore_skip_stat_unmatch(options);
3103        if (options->break_opt != -1)
3104                diffcore_break(options->break_opt);
3105        if (options->detect_rename)
3106                diffcore_rename(options);
3107        if (options->break_opt != -1)
3108                diffcore_merge_broken();
3109        if (options->pickaxe)
3110                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3111        if (options->orderfile)
3112                diffcore_order(options->orderfile);
3113        diff_resolve_rename_copy();
3114        diffcore_apply_filter(options->filter);
3115
3116        options->has_changes = !!diff_queued_diff.nr;
3117}
3118
3119
3120void diff_addremove(struct diff_options *options,
3121                    int addremove, unsigned mode,
3122                    const unsigned char *sha1,
3123                    const char *base, const char *path)
3124{
3125        char concatpath[PATH_MAX];
3126        struct diff_filespec *one, *two;
3127
3128        /* This may look odd, but it is a preparation for
3129         * feeding "there are unchanged files which should
3130         * not produce diffs, but when you are doing copy
3131         * detection you would need them, so here they are"
3132         * entries to the diff-core.  They will be prefixed
3133         * with something like '=' or '*' (I haven't decided
3134         * which but should not make any difference).
3135         * Feeding the same new and old to diff_change()
3136         * also has the same effect.
3137         * Before the final output happens, they are pruned after
3138         * merged into rename/copy pairs as appropriate.
3139         */
3140        if (options->reverse_diff)
3141                addremove = (addremove == '+' ? '-' :
3142                             addremove == '-' ? '+' : addremove);
3143
3144        if (!path) path = "";
3145        sprintf(concatpath, "%s%s", base, path);
3146        one = alloc_filespec(concatpath);
3147        two = alloc_filespec(concatpath);
3148
3149        if (addremove != '+')
3150                fill_filespec(one, sha1, mode);
3151        if (addremove != '-')
3152                fill_filespec(two, sha1, mode);
3153
3154        diff_queue(&diff_queued_diff, one, two);
3155        options->has_changes = 1;
3156}
3157
3158void diff_change(struct diff_options *options,
3159                 unsigned old_mode, unsigned new_mode,
3160                 const unsigned char *old_sha1,
3161                 const unsigned char *new_sha1,
3162                 const char *base, const char *path)
3163{
3164        char concatpath[PATH_MAX];
3165        struct diff_filespec *one, *two;
3166
3167        if (options->reverse_diff) {
3168                unsigned tmp;
3169                const unsigned char *tmp_c;
3170                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3171                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3172        }
3173        if (!path) path = "";
3174        sprintf(concatpath, "%s%s", base, path);
3175        one = alloc_filespec(concatpath);
3176        two = alloc_filespec(concatpath);
3177        fill_filespec(one, old_sha1, old_mode);
3178        fill_filespec(two, new_sha1, new_mode);
3179
3180        diff_queue(&diff_queued_diff, one, two);
3181        options->has_changes = 1;
3182}
3183
3184void diff_unmerge(struct diff_options *options,
3185                  const char *path,
3186                  unsigned mode, const unsigned char *sha1)
3187{
3188        struct diff_filespec *one, *two;
3189        one = alloc_filespec(path);
3190        two = alloc_filespec(path);
3191        fill_filespec(one, sha1, mode);
3192        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3193}