211235376de2166593945a646dcfa97405dc74d0
   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        /* The line is a newly added line.  Does it have funny leading
 506         * whitespaces?  In indent, SP should never precede a TAB.
 507         */
 508        for (i = col0; i < len; i++) {
 509                if (line[i] == '\t') {
 510                        last_tab_in_indent = i;
 511                        if ((whitespace_rule & WS_SPACE_BEFORE_TAB) &&
 512                            0 <= last_space_in_indent)
 513                                need_highlight_leading_space = 1;
 514                }
 515                else if (line[i] == ' ')
 516                        last_space_in_indent = i;
 517                else
 518                        break;
 519        }
 520        fputs(set, stdout);
 521        fwrite(line, col0, 1, stdout);
 522        fputs(reset, stdout);
 523        if (((i == len) || line[i] == '\n') && i != col0) {
 524                /* The whole line was indent */
 525                emit_line(ws, reset, line + col0, len - col0);
 526                return;
 527        }
 528        i = col0;
 529        if (need_highlight_leading_space) {
 530                while (i < last_tab_in_indent) {
 531                        if (line[i] == ' ') {
 532                                fputs(ws, stdout);
 533                                putchar(' ');
 534                                fputs(reset, stdout);
 535                        }
 536                        else
 537                                putchar(line[i]);
 538                        i++;
 539                }
 540        }
 541        tail = len - 1;
 542        if (line[tail] == '\n' && i < tail)
 543                tail--;
 544        if (whitespace_rule & WS_TRAILING_SPACE) {
 545                while (i < tail) {
 546                        if (!isspace(line[tail]))
 547                                break;
 548                        tail--;
 549                }
 550        }
 551        if ((i < tail && line[tail + 1] != '\n')) {
 552                /* This has whitespace between tail+1..len */
 553                fputs(set, stdout);
 554                fwrite(line + i, tail - i + 1, 1, stdout);
 555                fputs(reset, stdout);
 556                emit_line(ws, reset, line + tail + 1, len - tail - 1);
 557        }
 558        else
 559                emit_line(set, reset, line + i, len - i);
 560}
 561
 562static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 563{
 564        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 565        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 566
 567        if (!*ws)
 568                emit_line(set, reset, line, len);
 569        else
 570                emit_line_with_ws(ecbdata->nparents, set, reset, ws,
 571                                line, len);
 572}
 573
 574static void fn_out_consume(void *priv, char *line, unsigned long len)
 575{
 576        int i;
 577        int color;
 578        struct emit_callback *ecbdata = priv;
 579        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 580        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 581
 582        *(ecbdata->found_changesp) = 1;
 583
 584        if (ecbdata->label_path[0]) {
 585                const char *name_a_tab, *name_b_tab;
 586
 587                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 588                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 589
 590                printf("%s--- %s%s%s\n",
 591                       set, ecbdata->label_path[0], reset, name_a_tab);
 592                printf("%s+++ %s%s%s\n",
 593                       set, ecbdata->label_path[1], reset, name_b_tab);
 594                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 595        }
 596
 597        /* This is not really necessary for now because
 598         * this codepath only deals with two-way diffs.
 599         */
 600        for (i = 0; i < len && line[i] == '@'; i++)
 601                ;
 602        if (2 <= i && i < len && line[i] == ' ') {
 603                ecbdata->nparents = i - 1;
 604                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 605                          reset, line, len);
 606                return;
 607        }
 608
 609        if (len < ecbdata->nparents) {
 610                set = reset;
 611                emit_line(reset, reset, line, len);
 612                return;
 613        }
 614
 615        color = DIFF_PLAIN;
 616        if (ecbdata->diff_words && ecbdata->nparents != 1)
 617                /* fall back to normal diff */
 618                free_diff_words_data(ecbdata);
 619        if (ecbdata->diff_words) {
 620                if (line[0] == '-') {
 621                        diff_words_append(line, len,
 622                                          &ecbdata->diff_words->minus);
 623                        return;
 624                } else if (line[0] == '+') {
 625                        diff_words_append(line, len,
 626                                          &ecbdata->diff_words->plus);
 627                        return;
 628                }
 629                if (ecbdata->diff_words->minus.text.size ||
 630                    ecbdata->diff_words->plus.text.size)
 631                        diff_words_show(ecbdata->diff_words);
 632                line++;
 633                len--;
 634                emit_line(set, reset, line, len);
 635                return;
 636        }
 637        for (i = 0; i < ecbdata->nparents && len; i++) {
 638                if (line[i] == '-')
 639                        color = DIFF_FILE_OLD;
 640                else if (line[i] == '+')
 641                        color = DIFF_FILE_NEW;
 642        }
 643
 644        if (color != DIFF_FILE_NEW) {
 645                emit_line(diff_get_color(ecbdata->color_diff, color),
 646                          reset, line, len);
 647                return;
 648        }
 649        emit_add_line(reset, ecbdata, line, len);
 650}
 651
 652static char *pprint_rename(const char *a, const char *b)
 653{
 654        const char *old = a;
 655        const char *new = b;
 656        struct strbuf name;
 657        int pfx_length, sfx_length;
 658        int len_a = strlen(a);
 659        int len_b = strlen(b);
 660        int a_midlen, b_midlen;
 661        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 662        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 663
 664        strbuf_init(&name, 0);
 665        if (qlen_a || qlen_b) {
 666                quote_c_style(a, &name, NULL, 0);
 667                strbuf_addstr(&name, " => ");
 668                quote_c_style(b, &name, NULL, 0);
 669                return strbuf_detach(&name, NULL);
 670        }
 671
 672        /* Find common prefix */
 673        pfx_length = 0;
 674        while (*old && *new && *old == *new) {
 675                if (*old == '/')
 676                        pfx_length = old - a + 1;
 677                old++;
 678                new++;
 679        }
 680
 681        /* Find common suffix */
 682        old = a + len_a;
 683        new = b + len_b;
 684        sfx_length = 0;
 685        while (a <= old && b <= new && *old == *new) {
 686                if (*old == '/')
 687                        sfx_length = len_a - (old - a);
 688                old--;
 689                new--;
 690        }
 691
 692        /*
 693         * pfx{mid-a => mid-b}sfx
 694         * {pfx-a => pfx-b}sfx
 695         * pfx{sfx-a => sfx-b}
 696         * name-a => name-b
 697         */
 698        a_midlen = len_a - pfx_length - sfx_length;
 699        b_midlen = len_b - pfx_length - sfx_length;
 700        if (a_midlen < 0)
 701                a_midlen = 0;
 702        if (b_midlen < 0)
 703                b_midlen = 0;
 704
 705        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 706        if (pfx_length + sfx_length) {
 707                strbuf_add(&name, a, pfx_length);
 708                strbuf_addch(&name, '{');
 709        }
 710        strbuf_add(&name, a + pfx_length, a_midlen);
 711        strbuf_addstr(&name, " => ");
 712        strbuf_add(&name, b + pfx_length, b_midlen);
 713        if (pfx_length + sfx_length) {
 714                strbuf_addch(&name, '}');
 715                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 716        }
 717        return strbuf_detach(&name, NULL);
 718}
 719
 720struct diffstat_t {
 721        struct xdiff_emit_state xm;
 722
 723        int nr;
 724        int alloc;
 725        struct diffstat_file {
 726                char *name;
 727                unsigned is_unmerged:1;
 728                unsigned is_binary:1;
 729                unsigned is_renamed:1;
 730                unsigned int added, deleted;
 731        } **files;
 732};
 733
 734static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 735                                          const char *name_a,
 736                                          const char *name_b)
 737{
 738        struct diffstat_file *x;
 739        x = xcalloc(sizeof (*x), 1);
 740        if (diffstat->nr == diffstat->alloc) {
 741                diffstat->alloc = alloc_nr(diffstat->alloc);
 742                diffstat->files = xrealloc(diffstat->files,
 743                                diffstat->alloc * sizeof(x));
 744        }
 745        diffstat->files[diffstat->nr++] = x;
 746        if (name_b) {
 747                x->name = pprint_rename(name_a, name_b);
 748                x->is_renamed = 1;
 749        }
 750        else
 751                x->name = xstrdup(name_a);
 752        return x;
 753}
 754
 755static void diffstat_consume(void *priv, char *line, unsigned long len)
 756{
 757        struct diffstat_t *diffstat = priv;
 758        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 759
 760        if (line[0] == '+')
 761                x->added++;
 762        else if (line[0] == '-')
 763                x->deleted++;
 764}
 765
 766const char mime_boundary_leader[] = "------------";
 767
 768static int scale_linear(int it, int width, int max_change)
 769{
 770        /*
 771         * make sure that at least one '-' is printed if there were deletions,
 772         * and likewise for '+'.
 773         */
 774        if (max_change < 2)
 775                return it;
 776        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 777}
 778
 779static void show_name(const char *prefix, const char *name, int len,
 780                      const char *reset, const char *set)
 781{
 782        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 783}
 784
 785static void show_graph(char ch, int cnt, const char *set, const char *reset)
 786{
 787        if (cnt <= 0)
 788                return;
 789        printf("%s", set);
 790        while (cnt--)
 791                putchar(ch);
 792        printf("%s", reset);
 793}
 794
 795static void show_stats(struct diffstat_t* data, struct diff_options *options)
 796{
 797        int i, len, add, del, total, adds = 0, dels = 0;
 798        int max_change = 0, max_len = 0;
 799        int total_files = data->nr;
 800        int width, name_width;
 801        const char *reset, *set, *add_c, *del_c;
 802
 803        if (data->nr == 0)
 804                return;
 805
 806        width = options->stat_width ? options->stat_width : 80;
 807        name_width = options->stat_name_width ? options->stat_name_width : 50;
 808
 809        /* Sanity: give at least 5 columns to the graph,
 810         * but leave at least 10 columns for the name.
 811         */
 812        if (width < name_width + 15) {
 813                if (name_width <= 25)
 814                        width = name_width + 15;
 815                else
 816                        name_width = width - 15;
 817        }
 818
 819        /* Find the longest filename and max number of changes */
 820        reset = diff_get_color(options->color_diff, DIFF_RESET);
 821        set = diff_get_color(options->color_diff, DIFF_PLAIN);
 822        add_c = diff_get_color(options->color_diff, DIFF_FILE_NEW);
 823        del_c = diff_get_color(options->color_diff, DIFF_FILE_OLD);
 824
 825        for (i = 0; i < data->nr; i++) {
 826                struct diffstat_file *file = data->files[i];
 827                int change = file->added + file->deleted;
 828
 829                if (!file->is_renamed) {  /* renames are already quoted by pprint_rename */
 830                        struct strbuf buf;
 831                        strbuf_init(&buf, 0);
 832                        if (quote_c_style(file->name, &buf, NULL, 0)) {
 833                                free(file->name);
 834                                file->name = strbuf_detach(&buf, NULL);
 835                        } else {
 836                                strbuf_release(&buf);
 837                        }
 838                }
 839
 840                len = strlen(file->name);
 841                if (max_len < len)
 842                        max_len = len;
 843
 844                if (file->is_binary || file->is_unmerged)
 845                        continue;
 846                if (max_change < change)
 847                        max_change = change;
 848        }
 849
 850        /* Compute the width of the graph part;
 851         * 10 is for one blank at the beginning of the line plus
 852         * " | count " between the name and the graph.
 853         *
 854         * From here on, name_width is the width of the name area,
 855         * and width is the width of the graph area.
 856         */
 857        name_width = (name_width < max_len) ? name_width : max_len;
 858        if (width < (name_width + 10) + max_change)
 859                width = width - (name_width + 10);
 860        else
 861                width = max_change;
 862
 863        for (i = 0; i < data->nr; i++) {
 864                const char *prefix = "";
 865                char *name = data->files[i]->name;
 866                int added = data->files[i]->added;
 867                int deleted = data->files[i]->deleted;
 868                int name_len;
 869
 870                /*
 871                 * "scale" the filename
 872                 */
 873                len = name_width;
 874                name_len = strlen(name);
 875                if (name_width < name_len) {
 876                        char *slash;
 877                        prefix = "...";
 878                        len -= 3;
 879                        name += name_len - len;
 880                        slash = strchr(name, '/');
 881                        if (slash)
 882                                name = slash;
 883                }
 884
 885                if (data->files[i]->is_binary) {
 886                        show_name(prefix, name, len, reset, set);
 887                        printf("  Bin ");
 888                        printf("%s%d%s", del_c, deleted, reset);
 889                        printf(" -> ");
 890                        printf("%s%d%s", add_c, added, reset);
 891                        printf(" bytes");
 892                        printf("\n");
 893                        goto free_diffstat_file;
 894                }
 895                else if (data->files[i]->is_unmerged) {
 896                        show_name(prefix, name, len, reset, set);
 897                        printf("  Unmerged\n");
 898                        goto free_diffstat_file;
 899                }
 900                else if (!data->files[i]->is_renamed &&
 901                         (added + deleted == 0)) {
 902                        total_files--;
 903                        goto free_diffstat_file;
 904                }
 905
 906                /*
 907                 * scale the add/delete
 908                 */
 909                add = added;
 910                del = deleted;
 911                total = add + del;
 912                adds += add;
 913                dels += del;
 914
 915                if (width <= max_change) {
 916                        add = scale_linear(add, width, max_change);
 917                        del = scale_linear(del, width, max_change);
 918                        total = add + del;
 919                }
 920                show_name(prefix, name, len, reset, set);
 921                printf("%5d ", added + deleted);
 922                show_graph('+', add, add_c, reset);
 923                show_graph('-', del, del_c, reset);
 924                putchar('\n');
 925        free_diffstat_file:
 926                free(data->files[i]->name);
 927                free(data->files[i]);
 928        }
 929        free(data->files);
 930        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 931               set, total_files, adds, dels, reset);
 932}
 933
 934static void show_shortstats(struct diffstat_t* data)
 935{
 936        int i, adds = 0, dels = 0, total_files = data->nr;
 937
 938        if (data->nr == 0)
 939                return;
 940
 941        for (i = 0; i < data->nr; i++) {
 942                if (!data->files[i]->is_binary &&
 943                    !data->files[i]->is_unmerged) {
 944                        int added = data->files[i]->added;
 945                        int deleted= data->files[i]->deleted;
 946                        if (!data->files[i]->is_renamed &&
 947                            (added + deleted == 0)) {
 948                                total_files--;
 949                        } else {
 950                                adds += added;
 951                                dels += deleted;
 952                        }
 953                }
 954                free(data->files[i]->name);
 955                free(data->files[i]);
 956        }
 957        free(data->files);
 958
 959        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 960               total_files, adds, dels);
 961}
 962
 963static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 964{
 965        int i;
 966
 967        for (i = 0; i < data->nr; i++) {
 968                struct diffstat_file *file = data->files[i];
 969
 970                if (file->is_binary)
 971                        printf("-\t-\t");
 972                else
 973                        printf("%d\t%d\t", file->added, file->deleted);
 974                if (!file->is_renamed) {
 975                        write_name_quoted(file->name, stdout, options->line_termination);
 976                } else {
 977                        fputs(file->name, stdout);
 978                        putchar(options->line_termination);
 979                }
 980        }
 981}
 982
 983struct checkdiff_t {
 984        struct xdiff_emit_state xm;
 985        const char *filename;
 986        int lineno, color_diff;
 987};
 988
 989static void checkdiff_consume(void *priv, char *line, unsigned long len)
 990{
 991        struct checkdiff_t *data = priv;
 992        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
 993        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
 994        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
 995
 996        if (line[0] == '+') {
 997                int i, spaces = 0, space_before_tab = 0, white_space_at_end = 0;
 998
 999                /* check space before tab */
1000                for (i = 1; i < len && (line[i] == ' ' || line[i] == '\t'); i++)
1001                        if (line[i] == ' ')
1002                                spaces++;
1003                if (line[i - 1] == '\t' && spaces)
1004                        space_before_tab = 1;
1005
1006                /* check white space at line end */
1007                if (line[len - 1] == '\n')
1008                        len--;
1009                if (isspace(line[len - 1]))
1010                        white_space_at_end = 1;
1011
1012                if (space_before_tab || white_space_at_end) {
1013                        printf("%s:%d: %s", data->filename, data->lineno, ws);
1014                        if (space_before_tab) {
1015                                printf("space before tab");
1016                                if (white_space_at_end)
1017                                        putchar(',');
1018                        }
1019                        if (white_space_at_end)
1020                                printf("white space at end");
1021                        printf(":%s ", reset);
1022                        emit_line_with_ws(1, set, reset, ws, line, len);
1023                }
1024
1025                data->lineno++;
1026        } else if (line[0] == ' ')
1027                data->lineno++;
1028        else if (line[0] == '@') {
1029                char *plus = strchr(line, '+');
1030                if (plus)
1031                        data->lineno = strtol(plus, NULL, 10);
1032                else
1033                        die("invalid diff");
1034        }
1035}
1036
1037static unsigned char *deflate_it(char *data,
1038                                 unsigned long size,
1039                                 unsigned long *result_size)
1040{
1041        int bound;
1042        unsigned char *deflated;
1043        z_stream stream;
1044
1045        memset(&stream, 0, sizeof(stream));
1046        deflateInit(&stream, zlib_compression_level);
1047        bound = deflateBound(&stream, size);
1048        deflated = xmalloc(bound);
1049        stream.next_out = deflated;
1050        stream.avail_out = bound;
1051
1052        stream.next_in = (unsigned char *)data;
1053        stream.avail_in = size;
1054        while (deflate(&stream, Z_FINISH) == Z_OK)
1055                ; /* nothing */
1056        deflateEnd(&stream);
1057        *result_size = stream.total_out;
1058        return deflated;
1059}
1060
1061static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1062{
1063        void *cp;
1064        void *delta;
1065        void *deflated;
1066        void *data;
1067        unsigned long orig_size;
1068        unsigned long delta_size;
1069        unsigned long deflate_size;
1070        unsigned long data_size;
1071
1072        /* We could do deflated delta, or we could do just deflated two,
1073         * whichever is smaller.
1074         */
1075        delta = NULL;
1076        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1077        if (one->size && two->size) {
1078                delta = diff_delta(one->ptr, one->size,
1079                                   two->ptr, two->size,
1080                                   &delta_size, deflate_size);
1081                if (delta) {
1082                        void *to_free = delta;
1083                        orig_size = delta_size;
1084                        delta = deflate_it(delta, delta_size, &delta_size);
1085                        free(to_free);
1086                }
1087        }
1088
1089        if (delta && delta_size < deflate_size) {
1090                printf("delta %lu\n", orig_size);
1091                free(deflated);
1092                data = delta;
1093                data_size = delta_size;
1094        }
1095        else {
1096                printf("literal %lu\n", two->size);
1097                free(delta);
1098                data = deflated;
1099                data_size = deflate_size;
1100        }
1101
1102        /* emit data encoded in base85 */
1103        cp = data;
1104        while (data_size) {
1105                int bytes = (52 < data_size) ? 52 : data_size;
1106                char line[70];
1107                data_size -= bytes;
1108                if (bytes <= 26)
1109                        line[0] = bytes + 'A' - 1;
1110                else
1111                        line[0] = bytes - 26 + 'a' - 1;
1112                encode_85(line + 1, cp, bytes);
1113                cp = (char *) cp + bytes;
1114                puts(line);
1115        }
1116        printf("\n");
1117        free(data);
1118}
1119
1120static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1121{
1122        printf("GIT binary patch\n");
1123        emit_binary_diff_body(one, two);
1124        emit_binary_diff_body(two, one);
1125}
1126
1127static void setup_diff_attr_check(struct git_attr_check *check)
1128{
1129        static struct git_attr *attr_diff;
1130
1131        if (!attr_diff) {
1132                attr_diff = git_attr("diff", 4);
1133        }
1134        check[0].attr = attr_diff;
1135}
1136
1137static void diff_filespec_check_attr(struct diff_filespec *one)
1138{
1139        struct git_attr_check attr_diff_check;
1140        int check_from_data = 0;
1141
1142        if (one->checked_attr)
1143                return;
1144
1145        setup_diff_attr_check(&attr_diff_check);
1146        one->is_binary = 0;
1147        one->funcname_pattern_ident = NULL;
1148
1149        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1150                const char *value;
1151
1152                /* binaryness */
1153                value = attr_diff_check.value;
1154                if (ATTR_TRUE(value))
1155                        ;
1156                else if (ATTR_FALSE(value))
1157                        one->is_binary = 1;
1158                else
1159                        check_from_data = 1;
1160
1161                /* funcname pattern ident */
1162                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1163                        ;
1164                else
1165                        one->funcname_pattern_ident = value;
1166        }
1167
1168        if (check_from_data) {
1169                if (!one->data && DIFF_FILE_VALID(one))
1170                        diff_populate_filespec(one, 0);
1171
1172                if (one->data)
1173                        one->is_binary = buffer_is_binary(one->data, one->size);
1174        }
1175}
1176
1177int diff_filespec_is_binary(struct diff_filespec *one)
1178{
1179        diff_filespec_check_attr(one);
1180        return one->is_binary;
1181}
1182
1183static const char *funcname_pattern(const char *ident)
1184{
1185        struct funcname_pattern *pp;
1186
1187        read_config_if_needed();
1188        for (pp = funcname_pattern_list; pp; pp = pp->next)
1189                if (!strcmp(ident, pp->name))
1190                        return pp->pattern;
1191        return NULL;
1192}
1193
1194static struct builtin_funcname_pattern {
1195        const char *name;
1196        const char *pattern;
1197} builtin_funcname_pattern[] = {
1198        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1199                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1200                        "^[     ]*\\(\\([       ]*"
1201                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1202                        "[      ]*([^;]*$\\)" },
1203        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1204};
1205
1206static const char *diff_funcname_pattern(struct diff_filespec *one)
1207{
1208        const char *ident, *pattern;
1209        int i;
1210
1211        diff_filespec_check_attr(one);
1212        ident = one->funcname_pattern_ident;
1213
1214        if (!ident)
1215                /*
1216                 * If the config file has "funcname.default" defined, that
1217                 * regexp is used; otherwise NULL is returned and xemit uses
1218                 * the built-in default.
1219                 */
1220                return funcname_pattern("default");
1221
1222        /* Look up custom "funcname.$ident" regexp from config. */
1223        pattern = funcname_pattern(ident);
1224        if (pattern)
1225                return pattern;
1226
1227        /*
1228         * And define built-in fallback patterns here.  Note that
1229         * these can be overriden by the user's config settings.
1230         */
1231        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1232                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1233                        return builtin_funcname_pattern[i].pattern;
1234
1235        return NULL;
1236}
1237
1238static void builtin_diff(const char *name_a,
1239                         const char *name_b,
1240                         struct diff_filespec *one,
1241                         struct diff_filespec *two,
1242                         const char *xfrm_msg,
1243                         struct diff_options *o,
1244                         int complete_rewrite)
1245{
1246        mmfile_t mf1, mf2;
1247        const char *lbl[2];
1248        char *a_one, *b_two;
1249        const char *set = diff_get_color(o->color_diff, DIFF_METAINFO);
1250        const char *reset = diff_get_color(o->color_diff, DIFF_RESET);
1251
1252        a_one = quote_two("a/", name_a + (*name_a == '/'));
1253        b_two = quote_two("b/", name_b + (*name_b == '/'));
1254        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1255        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1256        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1257        if (lbl[0][0] == '/') {
1258                /* /dev/null */
1259                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1260                if (xfrm_msg && xfrm_msg[0])
1261                        printf("%s%s%s\n", set, xfrm_msg, reset);
1262        }
1263        else if (lbl[1][0] == '/') {
1264                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1265                if (xfrm_msg && xfrm_msg[0])
1266                        printf("%s%s%s\n", set, xfrm_msg, reset);
1267        }
1268        else {
1269                if (one->mode != two->mode) {
1270                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1271                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1272                }
1273                if (xfrm_msg && xfrm_msg[0])
1274                        printf("%s%s%s\n", set, xfrm_msg, reset);
1275                /*
1276                 * we do not run diff between different kind
1277                 * of objects.
1278                 */
1279                if ((one->mode ^ two->mode) & S_IFMT)
1280                        goto free_ab_and_return;
1281                if (complete_rewrite) {
1282                        emit_rewrite_diff(name_a, name_b, one, two,
1283                                        o->color_diff);
1284                        o->found_changes = 1;
1285                        goto free_ab_and_return;
1286                }
1287        }
1288
1289        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1290                die("unable to read files to diff");
1291
1292        if (!o->text &&
1293            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1294                /* Quite common confusing case */
1295                if (mf1.size == mf2.size &&
1296                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1297                        goto free_ab_and_return;
1298                if (o->binary)
1299                        emit_binary_diff(&mf1, &mf2);
1300                else
1301                        printf("Binary files %s and %s differ\n",
1302                               lbl[0], lbl[1]);
1303                o->found_changes = 1;
1304        }
1305        else {
1306                /* Crazy xdl interfaces.. */
1307                const char *diffopts = getenv("GIT_DIFF_OPTS");
1308                xpparam_t xpp;
1309                xdemitconf_t xecfg;
1310                xdemitcb_t ecb;
1311                struct emit_callback ecbdata;
1312                const char *funcname_pattern;
1313
1314                funcname_pattern = diff_funcname_pattern(one);
1315                if (!funcname_pattern)
1316                        funcname_pattern = diff_funcname_pattern(two);
1317
1318                memset(&xecfg, 0, sizeof(xecfg));
1319                memset(&ecbdata, 0, sizeof(ecbdata));
1320                ecbdata.label_path = lbl;
1321                ecbdata.color_diff = o->color_diff;
1322                ecbdata.found_changesp = &o->found_changes;
1323                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1324                xecfg.ctxlen = o->context;
1325                xecfg.flags = XDL_EMIT_FUNCNAMES;
1326                if (funcname_pattern)
1327                        xdiff_set_find_func(&xecfg, funcname_pattern);
1328                if (!diffopts)
1329                        ;
1330                else if (!prefixcmp(diffopts, "--unified="))
1331                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1332                else if (!prefixcmp(diffopts, "-u"))
1333                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1334                ecb.outf = xdiff_outf;
1335                ecb.priv = &ecbdata;
1336                ecbdata.xm.consume = fn_out_consume;
1337                if (o->color_diff_words)
1338                        ecbdata.diff_words =
1339                                xcalloc(1, sizeof(struct diff_words_data));
1340                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1341                if (o->color_diff_words)
1342                        free_diff_words_data(&ecbdata);
1343        }
1344
1345 free_ab_and_return:
1346        diff_free_filespec_data(one);
1347        diff_free_filespec_data(two);
1348        free(a_one);
1349        free(b_two);
1350        return;
1351}
1352
1353static void builtin_diffstat(const char *name_a, const char *name_b,
1354                             struct diff_filespec *one,
1355                             struct diff_filespec *two,
1356                             struct diffstat_t *diffstat,
1357                             struct diff_options *o,
1358                             int complete_rewrite)
1359{
1360        mmfile_t mf1, mf2;
1361        struct diffstat_file *data;
1362
1363        data = diffstat_add(diffstat, name_a, name_b);
1364
1365        if (!one || !two) {
1366                data->is_unmerged = 1;
1367                return;
1368        }
1369        if (complete_rewrite) {
1370                diff_populate_filespec(one, 0);
1371                diff_populate_filespec(two, 0);
1372                data->deleted = count_lines(one->data, one->size);
1373                data->added = count_lines(two->data, two->size);
1374                goto free_and_return;
1375        }
1376        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1377                die("unable to read files to diff");
1378
1379        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1380                data->is_binary = 1;
1381                data->added = mf2.size;
1382                data->deleted = mf1.size;
1383        } else {
1384                /* Crazy xdl interfaces.. */
1385                xpparam_t xpp;
1386                xdemitconf_t xecfg;
1387                xdemitcb_t ecb;
1388
1389                memset(&xecfg, 0, sizeof(xecfg));
1390                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1391                ecb.outf = xdiff_outf;
1392                ecb.priv = diffstat;
1393                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1394        }
1395
1396 free_and_return:
1397        diff_free_filespec_data(one);
1398        diff_free_filespec_data(two);
1399}
1400
1401static void builtin_checkdiff(const char *name_a, const char *name_b,
1402                             struct diff_filespec *one,
1403                             struct diff_filespec *two, struct diff_options *o)
1404{
1405        mmfile_t mf1, mf2;
1406        struct checkdiff_t data;
1407
1408        if (!two)
1409                return;
1410
1411        memset(&data, 0, sizeof(data));
1412        data.xm.consume = checkdiff_consume;
1413        data.filename = name_b ? name_b : name_a;
1414        data.lineno = 0;
1415        data.color_diff = o->color_diff;
1416
1417        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1418                die("unable to read files to diff");
1419
1420        if (diff_filespec_is_binary(two))
1421                goto free_and_return;
1422        else {
1423                /* Crazy xdl interfaces.. */
1424                xpparam_t xpp;
1425                xdemitconf_t xecfg;
1426                xdemitcb_t ecb;
1427
1428                memset(&xecfg, 0, sizeof(xecfg));
1429                xpp.flags = XDF_NEED_MINIMAL;
1430                ecb.outf = xdiff_outf;
1431                ecb.priv = &data;
1432                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1433        }
1434 free_and_return:
1435        diff_free_filespec_data(one);
1436        diff_free_filespec_data(two);
1437}
1438
1439struct diff_filespec *alloc_filespec(const char *path)
1440{
1441        int namelen = strlen(path);
1442        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1443
1444        memset(spec, 0, sizeof(*spec));
1445        spec->path = (char *)(spec + 1);
1446        memcpy(spec->path, path, namelen+1);
1447        spec->count = 1;
1448        return spec;
1449}
1450
1451void free_filespec(struct diff_filespec *spec)
1452{
1453        if (!--spec->count) {
1454                diff_free_filespec_data(spec);
1455                free(spec);
1456        }
1457}
1458
1459void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1460                   unsigned short mode)
1461{
1462        if (mode) {
1463                spec->mode = canon_mode(mode);
1464                hashcpy(spec->sha1, sha1);
1465                spec->sha1_valid = !is_null_sha1(sha1);
1466        }
1467}
1468
1469/*
1470 * Given a name and sha1 pair, if the index tells us the file in
1471 * the work tree has that object contents, return true, so that
1472 * prepare_temp_file() does not have to inflate and extract.
1473 */
1474static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1475{
1476        struct cache_entry *ce;
1477        struct stat st;
1478        int pos, len;
1479
1480        /* We do not read the cache ourselves here, because the
1481         * benchmark with my previous version that always reads cache
1482         * shows that it makes things worse for diff-tree comparing
1483         * two linux-2.6 kernel trees in an already checked out work
1484         * tree.  This is because most diff-tree comparisons deal with
1485         * only a small number of files, while reading the cache is
1486         * expensive for a large project, and its cost outweighs the
1487         * savings we get by not inflating the object to a temporary
1488         * file.  Practically, this code only helps when we are used
1489         * by diff-cache --cached, which does read the cache before
1490         * calling us.
1491         */
1492        if (!active_cache)
1493                return 0;
1494
1495        /* We want to avoid the working directory if our caller
1496         * doesn't need the data in a normal file, this system
1497         * is rather slow with its stat/open/mmap/close syscalls,
1498         * and the object is contained in a pack file.  The pack
1499         * is probably already open and will be faster to obtain
1500         * the data through than the working directory.  Loose
1501         * objects however would tend to be slower as they need
1502         * to be individually opened and inflated.
1503         */
1504        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1505                return 0;
1506
1507        len = strlen(name);
1508        pos = cache_name_pos(name, len);
1509        if (pos < 0)
1510                return 0;
1511        ce = active_cache[pos];
1512        if ((lstat(name, &st) < 0) ||
1513            !S_ISREG(st.st_mode) || /* careful! */
1514            ce_match_stat(ce, &st, 0) ||
1515            hashcmp(sha1, ce->sha1))
1516                return 0;
1517        /* we return 1 only when we can stat, it is a regular file,
1518         * stat information matches, and sha1 recorded in the cache
1519         * matches.  I.e. we know the file in the work tree really is
1520         * the same as the <name, sha1> pair.
1521         */
1522        return 1;
1523}
1524
1525static int populate_from_stdin(struct diff_filespec *s)
1526{
1527        struct strbuf buf;
1528        size_t size = 0;
1529
1530        strbuf_init(&buf, 0);
1531        if (strbuf_read(&buf, 0, 0) < 0)
1532                return error("error while reading from stdin %s",
1533                                     strerror(errno));
1534
1535        s->should_munmap = 0;
1536        s->data = strbuf_detach(&buf, &size);
1537        s->size = size;
1538        s->should_free = 1;
1539        return 0;
1540}
1541
1542static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1543{
1544        int len;
1545        char *data = xmalloc(100);
1546        len = snprintf(data, 100,
1547                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1548        s->data = data;
1549        s->size = len;
1550        s->should_free = 1;
1551        if (size_only) {
1552                s->data = NULL;
1553                free(data);
1554        }
1555        return 0;
1556}
1557
1558/*
1559 * While doing rename detection and pickaxe operation, we may need to
1560 * grab the data for the blob (or file) for our own in-core comparison.
1561 * diff_filespec has data and size fields for this purpose.
1562 */
1563int diff_populate_filespec(struct diff_filespec *s, int size_only)
1564{
1565        int err = 0;
1566        if (!DIFF_FILE_VALID(s))
1567                die("internal error: asking to populate invalid file.");
1568        if (S_ISDIR(s->mode))
1569                return -1;
1570
1571        if (s->data)
1572                return 0;
1573
1574        if (size_only && 0 < s->size)
1575                return 0;
1576
1577        if (S_ISGITLINK(s->mode))
1578                return diff_populate_gitlink(s, size_only);
1579
1580        if (!s->sha1_valid ||
1581            reuse_worktree_file(s->path, s->sha1, 0)) {
1582                struct strbuf buf;
1583                struct stat st;
1584                int fd;
1585
1586                if (!strcmp(s->path, "-"))
1587                        return populate_from_stdin(s);
1588
1589                if (lstat(s->path, &st) < 0) {
1590                        if (errno == ENOENT) {
1591                        err_empty:
1592                                err = -1;
1593                        empty:
1594                                s->data = (char *)"";
1595                                s->size = 0;
1596                                return err;
1597                        }
1598                }
1599                s->size = xsize_t(st.st_size);
1600                if (!s->size)
1601                        goto empty;
1602                if (size_only)
1603                        return 0;
1604                if (S_ISLNK(st.st_mode)) {
1605                        int ret;
1606                        s->data = xmalloc(s->size);
1607                        s->should_free = 1;
1608                        ret = readlink(s->path, s->data, s->size);
1609                        if (ret < 0) {
1610                                free(s->data);
1611                                goto err_empty;
1612                        }
1613                        return 0;
1614                }
1615                fd = open(s->path, O_RDONLY);
1616                if (fd < 0)
1617                        goto err_empty;
1618                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1619                close(fd);
1620                s->should_munmap = 1;
1621
1622                /*
1623                 * Convert from working tree format to canonical git format
1624                 */
1625                strbuf_init(&buf, 0);
1626                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1627                        size_t size = 0;
1628                        munmap(s->data, s->size);
1629                        s->should_munmap = 0;
1630                        s->data = strbuf_detach(&buf, &size);
1631                        s->size = size;
1632                        s->should_free = 1;
1633                }
1634        }
1635        else {
1636                enum object_type type;
1637                if (size_only)
1638                        type = sha1_object_info(s->sha1, &s->size);
1639                else {
1640                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1641                        s->should_free = 1;
1642                }
1643        }
1644        return 0;
1645}
1646
1647void diff_free_filespec_blob(struct diff_filespec *s)
1648{
1649        if (s->should_free)
1650                free(s->data);
1651        else if (s->should_munmap)
1652                munmap(s->data, s->size);
1653
1654        if (s->should_free || s->should_munmap) {
1655                s->should_free = s->should_munmap = 0;
1656                s->data = NULL;
1657        }
1658}
1659
1660void diff_free_filespec_data(struct diff_filespec *s)
1661{
1662        diff_free_filespec_blob(s);
1663        free(s->cnt_data);
1664        s->cnt_data = NULL;
1665}
1666
1667static void prep_temp_blob(struct diff_tempfile *temp,
1668                           void *blob,
1669                           unsigned long size,
1670                           const unsigned char *sha1,
1671                           int mode)
1672{
1673        int fd;
1674
1675        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1676        if (fd < 0)
1677                die("unable to create temp-file: %s", strerror(errno));
1678        if (write_in_full(fd, blob, size) != size)
1679                die("unable to write temp-file");
1680        close(fd);
1681        temp->name = temp->tmp_path;
1682        strcpy(temp->hex, sha1_to_hex(sha1));
1683        temp->hex[40] = 0;
1684        sprintf(temp->mode, "%06o", mode);
1685}
1686
1687static void prepare_temp_file(const char *name,
1688                              struct diff_tempfile *temp,
1689                              struct diff_filespec *one)
1690{
1691        if (!DIFF_FILE_VALID(one)) {
1692        not_a_valid_file:
1693                /* A '-' entry produces this for file-2, and
1694                 * a '+' entry produces this for file-1.
1695                 */
1696                temp->name = "/dev/null";
1697                strcpy(temp->hex, ".");
1698                strcpy(temp->mode, ".");
1699                return;
1700        }
1701
1702        if (!one->sha1_valid ||
1703            reuse_worktree_file(name, one->sha1, 1)) {
1704                struct stat st;
1705                if (lstat(name, &st) < 0) {
1706                        if (errno == ENOENT)
1707                                goto not_a_valid_file;
1708                        die("stat(%s): %s", name, strerror(errno));
1709                }
1710                if (S_ISLNK(st.st_mode)) {
1711                        int ret;
1712                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1713                        size_t sz = xsize_t(st.st_size);
1714                        if (sizeof(buf) <= st.st_size)
1715                                die("symlink too long: %s", name);
1716                        ret = readlink(name, buf, sz);
1717                        if (ret < 0)
1718                                die("readlink(%s)", name);
1719                        prep_temp_blob(temp, buf, sz,
1720                                       (one->sha1_valid ?
1721                                        one->sha1 : null_sha1),
1722                                       (one->sha1_valid ?
1723                                        one->mode : S_IFLNK));
1724                }
1725                else {
1726                        /* we can borrow from the file in the work tree */
1727                        temp->name = name;
1728                        if (!one->sha1_valid)
1729                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1730                        else
1731                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1732                        /* Even though we may sometimes borrow the
1733                         * contents from the work tree, we always want
1734                         * one->mode.  mode is trustworthy even when
1735                         * !(one->sha1_valid), as long as
1736                         * DIFF_FILE_VALID(one).
1737                         */
1738                        sprintf(temp->mode, "%06o", one->mode);
1739                }
1740                return;
1741        }
1742        else {
1743                if (diff_populate_filespec(one, 0))
1744                        die("cannot read data blob for %s", one->path);
1745                prep_temp_blob(temp, one->data, one->size,
1746                               one->sha1, one->mode);
1747        }
1748}
1749
1750static void remove_tempfile(void)
1751{
1752        int i;
1753
1754        for (i = 0; i < 2; i++)
1755                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1756                        unlink(diff_temp[i].name);
1757                        diff_temp[i].name = NULL;
1758                }
1759}
1760
1761static void remove_tempfile_on_signal(int signo)
1762{
1763        remove_tempfile();
1764        signal(SIGINT, SIG_DFL);
1765        raise(signo);
1766}
1767
1768/* An external diff command takes:
1769 *
1770 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1771 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1772 *
1773 */
1774static void run_external_diff(const char *pgm,
1775                              const char *name,
1776                              const char *other,
1777                              struct diff_filespec *one,
1778                              struct diff_filespec *two,
1779                              const char *xfrm_msg,
1780                              int complete_rewrite)
1781{
1782        const char *spawn_arg[10];
1783        struct diff_tempfile *temp = diff_temp;
1784        int retval;
1785        static int atexit_asked = 0;
1786        const char *othername;
1787        const char **arg = &spawn_arg[0];
1788
1789        othername = (other? other : name);
1790        if (one && two) {
1791                prepare_temp_file(name, &temp[0], one);
1792                prepare_temp_file(othername, &temp[1], two);
1793                if (! atexit_asked &&
1794                    (temp[0].name == temp[0].tmp_path ||
1795                     temp[1].name == temp[1].tmp_path)) {
1796                        atexit_asked = 1;
1797                        atexit(remove_tempfile);
1798                }
1799                signal(SIGINT, remove_tempfile_on_signal);
1800        }
1801
1802        if (one && two) {
1803                *arg++ = pgm;
1804                *arg++ = name;
1805                *arg++ = temp[0].name;
1806                *arg++ = temp[0].hex;
1807                *arg++ = temp[0].mode;
1808                *arg++ = temp[1].name;
1809                *arg++ = temp[1].hex;
1810                *arg++ = temp[1].mode;
1811                if (other) {
1812                        *arg++ = other;
1813                        *arg++ = xfrm_msg;
1814                }
1815        } else {
1816                *arg++ = pgm;
1817                *arg++ = name;
1818        }
1819        *arg = NULL;
1820        fflush(NULL);
1821        retval = run_command_v_opt(spawn_arg, 0);
1822        remove_tempfile();
1823        if (retval) {
1824                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1825                exit(1);
1826        }
1827}
1828
1829static const char *external_diff_attr(const char *name)
1830{
1831        struct git_attr_check attr_diff_check;
1832
1833        setup_diff_attr_check(&attr_diff_check);
1834        if (!git_checkattr(name, 1, &attr_diff_check)) {
1835                const char *value = attr_diff_check.value;
1836                if (!ATTR_TRUE(value) &&
1837                    !ATTR_FALSE(value) &&
1838                    !ATTR_UNSET(value)) {
1839                        struct ll_diff_driver *drv;
1840
1841                        read_config_if_needed();
1842                        for (drv = user_diff; drv; drv = drv->next)
1843                                if (!strcmp(drv->name, value))
1844                                        return drv->cmd;
1845                }
1846        }
1847        return NULL;
1848}
1849
1850static void run_diff_cmd(const char *pgm,
1851                         const char *name,
1852                         const char *other,
1853                         struct diff_filespec *one,
1854                         struct diff_filespec *two,
1855                         const char *xfrm_msg,
1856                         struct diff_options *o,
1857                         int complete_rewrite)
1858{
1859        if (!o->allow_external)
1860                pgm = NULL;
1861        else {
1862                const char *cmd = external_diff_attr(name);
1863                if (cmd)
1864                        pgm = cmd;
1865        }
1866
1867        if (pgm) {
1868                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1869                                  complete_rewrite);
1870                return;
1871        }
1872        if (one && two)
1873                builtin_diff(name, other ? other : name,
1874                             one, two, xfrm_msg, o, complete_rewrite);
1875        else
1876                printf("* Unmerged path %s\n", name);
1877}
1878
1879static void diff_fill_sha1_info(struct diff_filespec *one)
1880{
1881        if (DIFF_FILE_VALID(one)) {
1882                if (!one->sha1_valid) {
1883                        struct stat st;
1884                        if (!strcmp(one->path, "-")) {
1885                                hashcpy(one->sha1, null_sha1);
1886                                return;
1887                        }
1888                        if (lstat(one->path, &st) < 0)
1889                                die("stat %s", one->path);
1890                        if (index_path(one->sha1, one->path, &st, 0))
1891                                die("cannot hash %s\n", one->path);
1892                }
1893        }
1894        else
1895                hashclr(one->sha1);
1896}
1897
1898static int similarity_index(struct diff_filepair *p)
1899{
1900        return p->score * 100 / MAX_SCORE;
1901}
1902
1903static void run_diff(struct diff_filepair *p, struct diff_options *o)
1904{
1905        const char *pgm = external_diff();
1906        struct strbuf msg;
1907        char *xfrm_msg;
1908        struct diff_filespec *one = p->one;
1909        struct diff_filespec *two = p->two;
1910        const char *name;
1911        const char *other;
1912        int complete_rewrite = 0;
1913
1914
1915        if (DIFF_PAIR_UNMERGED(p)) {
1916                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1917                return;
1918        }
1919
1920        name  = p->one->path;
1921        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1922        diff_fill_sha1_info(one);
1923        diff_fill_sha1_info(two);
1924
1925        strbuf_init(&msg, PATH_MAX * 2 + 300);
1926        switch (p->status) {
1927        case DIFF_STATUS_COPIED:
1928                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1929                strbuf_addstr(&msg, "\ncopy from ");
1930                quote_c_style(name, &msg, NULL, 0);
1931                strbuf_addstr(&msg, "\ncopy to ");
1932                quote_c_style(other, &msg, NULL, 0);
1933                strbuf_addch(&msg, '\n');
1934                break;
1935        case DIFF_STATUS_RENAMED:
1936                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1937                strbuf_addstr(&msg, "\nrename from ");
1938                quote_c_style(name, &msg, NULL, 0);
1939                strbuf_addstr(&msg, "\nrename to ");
1940                quote_c_style(other, &msg, NULL, 0);
1941                strbuf_addch(&msg, '\n');
1942                break;
1943        case DIFF_STATUS_MODIFIED:
1944                if (p->score) {
1945                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
1946                                        similarity_index(p));
1947                        complete_rewrite = 1;
1948                        break;
1949                }
1950                /* fallthru */
1951        default:
1952                /* nothing */
1953                ;
1954        }
1955
1956        if (hashcmp(one->sha1, two->sha1)) {
1957                int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
1958
1959                if (o->binary) {
1960                        mmfile_t mf;
1961                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
1962                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
1963                                abbrev = 40;
1964                }
1965                strbuf_addf(&msg, "index %.*s..%.*s",
1966                                abbrev, sha1_to_hex(one->sha1),
1967                                abbrev, sha1_to_hex(two->sha1));
1968                if (one->mode == two->mode)
1969                        strbuf_addf(&msg, " %06o", one->mode);
1970                strbuf_addch(&msg, '\n');
1971        }
1972
1973        if (msg.len)
1974                strbuf_setlen(&msg, msg.len - 1);
1975        xfrm_msg = msg.len ? msg.buf : NULL;
1976
1977        if (!pgm &&
1978            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1979            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1980                /* a filepair that changes between file and symlink
1981                 * needs to be split into deletion and creation.
1982                 */
1983                struct diff_filespec *null = alloc_filespec(two->path);
1984                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1985                free(null);
1986                null = alloc_filespec(one->path);
1987                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1988                free(null);
1989        }
1990        else
1991                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
1992                             complete_rewrite);
1993
1994        strbuf_release(&msg);
1995}
1996
1997static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
1998                         struct diffstat_t *diffstat)
1999{
2000        const char *name;
2001        const char *other;
2002        int complete_rewrite = 0;
2003
2004        if (DIFF_PAIR_UNMERGED(p)) {
2005                /* unmerged */
2006                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2007                return;
2008        }
2009
2010        name = p->one->path;
2011        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2012
2013        diff_fill_sha1_info(p->one);
2014        diff_fill_sha1_info(p->two);
2015
2016        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2017                complete_rewrite = 1;
2018        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2019}
2020
2021static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2022{
2023        const char *name;
2024        const char *other;
2025
2026        if (DIFF_PAIR_UNMERGED(p)) {
2027                /* unmerged */
2028                return;
2029        }
2030
2031        name = p->one->path;
2032        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2033
2034        diff_fill_sha1_info(p->one);
2035        diff_fill_sha1_info(p->two);
2036
2037        builtin_checkdiff(name, other, p->one, p->two, o);
2038}
2039
2040void diff_setup(struct diff_options *options)
2041{
2042        memset(options, 0, sizeof(*options));
2043        options->line_termination = '\n';
2044        options->break_opt = -1;
2045        options->rename_limit = -1;
2046        options->context = 3;
2047        options->msg_sep = "";
2048
2049        options->change = diff_change;
2050        options->add_remove = diff_addremove;
2051        options->color_diff = diff_use_color_default;
2052        options->detect_rename = diff_detect_rename_default;
2053}
2054
2055int diff_setup_done(struct diff_options *options)
2056{
2057        int count = 0;
2058
2059        if (options->output_format & DIFF_FORMAT_NAME)
2060                count++;
2061        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2062                count++;
2063        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2064                count++;
2065        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2066                count++;
2067        if (count > 1)
2068                die("--name-only, --name-status, --check and -s are mutually exclusive");
2069
2070        if (options->find_copies_harder)
2071                options->detect_rename = DIFF_DETECT_COPY;
2072
2073        if (options->output_format & (DIFF_FORMAT_NAME |
2074                                      DIFF_FORMAT_NAME_STATUS |
2075                                      DIFF_FORMAT_CHECKDIFF |
2076                                      DIFF_FORMAT_NO_OUTPUT))
2077                options->output_format &= ~(DIFF_FORMAT_RAW |
2078                                            DIFF_FORMAT_NUMSTAT |
2079                                            DIFF_FORMAT_DIFFSTAT |
2080                                            DIFF_FORMAT_SHORTSTAT |
2081                                            DIFF_FORMAT_SUMMARY |
2082                                            DIFF_FORMAT_PATCH);
2083
2084        /*
2085         * These cases always need recursive; we do not drop caller-supplied
2086         * recursive bits for other formats here.
2087         */
2088        if (options->output_format & (DIFF_FORMAT_PATCH |
2089                                      DIFF_FORMAT_NUMSTAT |
2090                                      DIFF_FORMAT_DIFFSTAT |
2091                                      DIFF_FORMAT_SHORTSTAT |
2092                                      DIFF_FORMAT_SUMMARY |
2093                                      DIFF_FORMAT_CHECKDIFF))
2094                options->recursive = 1;
2095        /*
2096         * Also pickaxe would not work very well if you do not say recursive
2097         */
2098        if (options->pickaxe)
2099                options->recursive = 1;
2100
2101        if (options->detect_rename && options->rename_limit < 0)
2102                options->rename_limit = diff_rename_limit_default;
2103        if (options->setup & DIFF_SETUP_USE_CACHE) {
2104                if (!active_cache)
2105                        /* read-cache does not die even when it fails
2106                         * so it is safe for us to do this here.  Also
2107                         * it does not smudge active_cache or active_nr
2108                         * when it fails, so we do not have to worry about
2109                         * cleaning it up ourselves either.
2110                         */
2111                        read_cache();
2112        }
2113        if (options->abbrev <= 0 || 40 < options->abbrev)
2114                options->abbrev = 40; /* full */
2115
2116        /*
2117         * It does not make sense to show the first hit we happened
2118         * to have found.  It does not make sense not to return with
2119         * exit code in such a case either.
2120         */
2121        if (options->quiet) {
2122                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2123                options->exit_with_status = 1;
2124        }
2125
2126        /*
2127         * If we postprocess in diffcore, we cannot simply return
2128         * upon the first hit.  We need to run diff as usual.
2129         */
2130        if (options->pickaxe || options->filter)
2131                options->quiet = 0;
2132
2133        return 0;
2134}
2135
2136static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2137{
2138        char c, *eq;
2139        int len;
2140
2141        if (*arg != '-')
2142                return 0;
2143        c = *++arg;
2144        if (!c)
2145                return 0;
2146        if (c == arg_short) {
2147                c = *++arg;
2148                if (!c)
2149                        return 1;
2150                if (val && isdigit(c)) {
2151                        char *end;
2152                        int n = strtoul(arg, &end, 10);
2153                        if (*end)
2154                                return 0;
2155                        *val = n;
2156                        return 1;
2157                }
2158                return 0;
2159        }
2160        if (c != '-')
2161                return 0;
2162        arg++;
2163        eq = strchr(arg, '=');
2164        if (eq)
2165                len = eq - arg;
2166        else
2167                len = strlen(arg);
2168        if (!len || strncmp(arg, arg_long, len))
2169                return 0;
2170        if (eq) {
2171                int n;
2172                char *end;
2173                if (!isdigit(*++eq))
2174                        return 0;
2175                n = strtoul(eq, &end, 10);
2176                if (*end)
2177                        return 0;
2178                *val = n;
2179        }
2180        return 1;
2181}
2182
2183static int diff_scoreopt_parse(const char *opt);
2184
2185int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2186{
2187        const char *arg = av[0];
2188        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2189                options->output_format |= DIFF_FORMAT_PATCH;
2190        else if (opt_arg(arg, 'U', "unified", &options->context))
2191                options->output_format |= DIFF_FORMAT_PATCH;
2192        else if (!strcmp(arg, "--raw"))
2193                options->output_format |= DIFF_FORMAT_RAW;
2194        else if (!strcmp(arg, "--patch-with-raw")) {
2195                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2196        }
2197        else if (!strcmp(arg, "--numstat")) {
2198                options->output_format |= DIFF_FORMAT_NUMSTAT;
2199        }
2200        else if (!strcmp(arg, "--shortstat")) {
2201                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2202        }
2203        else if (!prefixcmp(arg, "--stat")) {
2204                char *end;
2205                int width = options->stat_width;
2206                int name_width = options->stat_name_width;
2207                arg += 6;
2208                end = (char *)arg;
2209
2210                switch (*arg) {
2211                case '-':
2212                        if (!prefixcmp(arg, "-width="))
2213                                width = strtoul(arg + 7, &end, 10);
2214                        else if (!prefixcmp(arg, "-name-width="))
2215                                name_width = strtoul(arg + 12, &end, 10);
2216                        break;
2217                case '=':
2218                        width = strtoul(arg+1, &end, 10);
2219                        if (*end == ',')
2220                                name_width = strtoul(end+1, &end, 10);
2221                }
2222
2223                /* Important! This checks all the error cases! */
2224                if (*end)
2225                        return 0;
2226                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2227                options->stat_name_width = name_width;
2228                options->stat_width = width;
2229        }
2230        else if (!strcmp(arg, "--check"))
2231                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2232        else if (!strcmp(arg, "--summary"))
2233                options->output_format |= DIFF_FORMAT_SUMMARY;
2234        else if (!strcmp(arg, "--patch-with-stat")) {
2235                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2236        }
2237        else if (!strcmp(arg, "-z"))
2238                options->line_termination = 0;
2239        else if (!prefixcmp(arg, "-l"))
2240                options->rename_limit = strtoul(arg+2, NULL, 10);
2241        else if (!strcmp(arg, "--full-index"))
2242                options->full_index = 1;
2243        else if (!strcmp(arg, "--binary")) {
2244                options->output_format |= DIFF_FORMAT_PATCH;
2245                options->binary = 1;
2246        }
2247        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text")) {
2248                options->text = 1;
2249        }
2250        else if (!strcmp(arg, "--name-only"))
2251                options->output_format |= DIFF_FORMAT_NAME;
2252        else if (!strcmp(arg, "--name-status"))
2253                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2254        else if (!strcmp(arg, "-R"))
2255                options->reverse_diff = 1;
2256        else if (!prefixcmp(arg, "-S"))
2257                options->pickaxe = arg + 2;
2258        else if (!strcmp(arg, "-s")) {
2259                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2260        }
2261        else if (!prefixcmp(arg, "-O"))
2262                options->orderfile = arg + 2;
2263        else if (!prefixcmp(arg, "--diff-filter="))
2264                options->filter = arg + 14;
2265        else if (!strcmp(arg, "--pickaxe-all"))
2266                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2267        else if (!strcmp(arg, "--pickaxe-regex"))
2268                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2269        else if (!prefixcmp(arg, "-B")) {
2270                if ((options->break_opt =
2271                     diff_scoreopt_parse(arg)) == -1)
2272                        return -1;
2273        }
2274        else if (!prefixcmp(arg, "-M")) {
2275                if ((options->rename_score =
2276                     diff_scoreopt_parse(arg)) == -1)
2277                        return -1;
2278                options->detect_rename = DIFF_DETECT_RENAME;
2279        }
2280        else if (!prefixcmp(arg, "-C")) {
2281                if (options->detect_rename == DIFF_DETECT_COPY)
2282                        options->find_copies_harder = 1;
2283                if ((options->rename_score =
2284                     diff_scoreopt_parse(arg)) == -1)
2285                        return -1;
2286                options->detect_rename = DIFF_DETECT_COPY;
2287        }
2288        else if (!strcmp(arg, "--find-copies-harder"))
2289                options->find_copies_harder = 1;
2290        else if (!strcmp(arg, "--follow"))
2291                options->follow_renames = 1;
2292        else if (!strcmp(arg, "--abbrev"))
2293                options->abbrev = DEFAULT_ABBREV;
2294        else if (!prefixcmp(arg, "--abbrev=")) {
2295                options->abbrev = strtoul(arg + 9, NULL, 10);
2296                if (options->abbrev < MINIMUM_ABBREV)
2297                        options->abbrev = MINIMUM_ABBREV;
2298                else if (40 < options->abbrev)
2299                        options->abbrev = 40;
2300        }
2301        else if (!strcmp(arg, "--color"))
2302                options->color_diff = 1;
2303        else if (!strcmp(arg, "--no-color"))
2304                options->color_diff = 0;
2305        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2306                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2307        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2308                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2309        else if (!strcmp(arg, "--ignore-space-at-eol"))
2310                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2311        else if (!strcmp(arg, "--color-words"))
2312                options->color_diff = options->color_diff_words = 1;
2313        else if (!strcmp(arg, "--no-renames"))
2314                options->detect_rename = 0;
2315        else if (!strcmp(arg, "--exit-code"))
2316                options->exit_with_status = 1;
2317        else if (!strcmp(arg, "--quiet"))
2318                options->quiet = 1;
2319        else if (!strcmp(arg, "--ext-diff"))
2320                options->allow_external = 1;
2321        else if (!strcmp(arg, "--no-ext-diff"))
2322                options->allow_external = 0;
2323        else
2324                return 0;
2325        return 1;
2326}
2327
2328static int parse_num(const char **cp_p)
2329{
2330        unsigned long num, scale;
2331        int ch, dot;
2332        const char *cp = *cp_p;
2333
2334        num = 0;
2335        scale = 1;
2336        dot = 0;
2337        for(;;) {
2338                ch = *cp;
2339                if ( !dot && ch == '.' ) {
2340                        scale = 1;
2341                        dot = 1;
2342                } else if ( ch == '%' ) {
2343                        scale = dot ? scale*100 : 100;
2344                        cp++;   /* % is always at the end */
2345                        break;
2346                } else if ( ch >= '0' && ch <= '9' ) {
2347                        if ( scale < 100000 ) {
2348                                scale *= 10;
2349                                num = (num*10) + (ch-'0');
2350                        }
2351                } else {
2352                        break;
2353                }
2354                cp++;
2355        }
2356        *cp_p = cp;
2357
2358        /* user says num divided by scale and we say internally that
2359         * is MAX_SCORE * num / scale.
2360         */
2361        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2362}
2363
2364static int diff_scoreopt_parse(const char *opt)
2365{
2366        int opt1, opt2, cmd;
2367
2368        if (*opt++ != '-')
2369                return -1;
2370        cmd = *opt++;
2371        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2372                return -1; /* that is not a -M, -C nor -B option */
2373
2374        opt1 = parse_num(&opt);
2375        if (cmd != 'B')
2376                opt2 = 0;
2377        else {
2378                if (*opt == 0)
2379                        opt2 = 0;
2380                else if (*opt != '/')
2381                        return -1; /* we expect -B80/99 or -B80 */
2382                else {
2383                        opt++;
2384                        opt2 = parse_num(&opt);
2385                }
2386        }
2387        if (*opt != 0)
2388                return -1;
2389        return opt1 | (opt2 << 16);
2390}
2391
2392struct diff_queue_struct diff_queued_diff;
2393
2394void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2395{
2396        if (queue->alloc <= queue->nr) {
2397                queue->alloc = alloc_nr(queue->alloc);
2398                queue->queue = xrealloc(queue->queue,
2399                                        sizeof(dp) * queue->alloc);
2400        }
2401        queue->queue[queue->nr++] = dp;
2402}
2403
2404struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2405                                 struct diff_filespec *one,
2406                                 struct diff_filespec *two)
2407{
2408        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2409        dp->one = one;
2410        dp->two = two;
2411        if (queue)
2412                diff_q(queue, dp);
2413        return dp;
2414}
2415
2416void diff_free_filepair(struct diff_filepair *p)
2417{
2418        free_filespec(p->one);
2419        free_filespec(p->two);
2420        free(p);
2421}
2422
2423/* This is different from find_unique_abbrev() in that
2424 * it stuffs the result with dots for alignment.
2425 */
2426const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2427{
2428        int abblen;
2429        const char *abbrev;
2430        if (len == 40)
2431                return sha1_to_hex(sha1);
2432
2433        abbrev = find_unique_abbrev(sha1, len);
2434        if (!abbrev)
2435                return sha1_to_hex(sha1);
2436        abblen = strlen(abbrev);
2437        if (abblen < 37) {
2438                static char hex[41];
2439                if (len < abblen && abblen <= len + 2)
2440                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2441                else
2442                        sprintf(hex, "%s...", abbrev);
2443                return hex;
2444        }
2445        return sha1_to_hex(sha1);
2446}
2447
2448static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2449{
2450        int line_termination = opt->line_termination;
2451        int inter_name_termination = line_termination ? '\t' : '\0';
2452
2453        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2454                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2455                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2456                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2457        }
2458        if (p->score) {
2459                printf("%c%03d%c", p->status, similarity_index(p),
2460                           inter_name_termination);
2461        } else {
2462                printf("%c%c", p->status, inter_name_termination);
2463        }
2464
2465        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2466                write_name_quoted(p->one->path, stdout, inter_name_termination);
2467                write_name_quoted(p->two->path, stdout, line_termination);
2468        } else {
2469                const char *path = p->one->mode ? p->one->path : p->two->path;
2470                write_name_quoted(path, stdout, line_termination);
2471        }
2472}
2473
2474int diff_unmodified_pair(struct diff_filepair *p)
2475{
2476        /* This function is written stricter than necessary to support
2477         * the currently implemented transformers, but the idea is to
2478         * let transformers to produce diff_filepairs any way they want,
2479         * and filter and clean them up here before producing the output.
2480         */
2481        struct diff_filespec *one = p->one, *two = p->two;
2482
2483        if (DIFF_PAIR_UNMERGED(p))
2484                return 0; /* unmerged is interesting */
2485
2486        /* deletion, addition, mode or type change
2487         * and rename are all interesting.
2488         */
2489        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2490            DIFF_PAIR_MODE_CHANGED(p) ||
2491            strcmp(one->path, two->path))
2492                return 0;
2493
2494        /* both are valid and point at the same path.  that is, we are
2495         * dealing with a change.
2496         */
2497        if (one->sha1_valid && two->sha1_valid &&
2498            !hashcmp(one->sha1, two->sha1))
2499                return 1; /* no change */
2500        if (!one->sha1_valid && !two->sha1_valid)
2501                return 1; /* both look at the same file on the filesystem. */
2502        return 0;
2503}
2504
2505static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2506{
2507        if (diff_unmodified_pair(p))
2508                return;
2509
2510        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2511            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2512                return; /* no tree diffs in patch format */
2513
2514        run_diff(p, o);
2515}
2516
2517static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2518                            struct diffstat_t *diffstat)
2519{
2520        if (diff_unmodified_pair(p))
2521                return;
2522
2523        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2524            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2525                return; /* no tree diffs in patch format */
2526
2527        run_diffstat(p, o, diffstat);
2528}
2529
2530static void diff_flush_checkdiff(struct diff_filepair *p,
2531                struct diff_options *o)
2532{
2533        if (diff_unmodified_pair(p))
2534                return;
2535
2536        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2537            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2538                return; /* no tree diffs in patch format */
2539
2540        run_checkdiff(p, o);
2541}
2542
2543int diff_queue_is_empty(void)
2544{
2545        struct diff_queue_struct *q = &diff_queued_diff;
2546        int i;
2547        for (i = 0; i < q->nr; i++)
2548                if (!diff_unmodified_pair(q->queue[i]))
2549                        return 0;
2550        return 1;
2551}
2552
2553#if DIFF_DEBUG
2554void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2555{
2556        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2557                x, one ? one : "",
2558                s->path,
2559                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2560                s->mode,
2561                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2562        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2563                x, one ? one : "",
2564                s->size, s->xfrm_flags);
2565}
2566
2567void diff_debug_filepair(const struct diff_filepair *p, int i)
2568{
2569        diff_debug_filespec(p->one, i, "one");
2570        diff_debug_filespec(p->two, i, "two");
2571        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2572                p->score, p->status ? p->status : '?',
2573                p->one->rename_used, p->broken_pair);
2574}
2575
2576void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2577{
2578        int i;
2579        if (msg)
2580                fprintf(stderr, "%s\n", msg);
2581        fprintf(stderr, "q->nr = %d\n", q->nr);
2582        for (i = 0; i < q->nr; i++) {
2583                struct diff_filepair *p = q->queue[i];
2584                diff_debug_filepair(p, i);
2585        }
2586}
2587#endif
2588
2589static void diff_resolve_rename_copy(void)
2590{
2591        int i;
2592        struct diff_filepair *p;
2593        struct diff_queue_struct *q = &diff_queued_diff;
2594
2595        diff_debug_queue("resolve-rename-copy", q);
2596
2597        for (i = 0; i < q->nr; i++) {
2598                p = q->queue[i];
2599                p->status = 0; /* undecided */
2600                if (DIFF_PAIR_UNMERGED(p))
2601                        p->status = DIFF_STATUS_UNMERGED;
2602                else if (!DIFF_FILE_VALID(p->one))
2603                        p->status = DIFF_STATUS_ADDED;
2604                else if (!DIFF_FILE_VALID(p->two))
2605                        p->status = DIFF_STATUS_DELETED;
2606                else if (DIFF_PAIR_TYPE_CHANGED(p))
2607                        p->status = DIFF_STATUS_TYPE_CHANGED;
2608
2609                /* from this point on, we are dealing with a pair
2610                 * whose both sides are valid and of the same type, i.e.
2611                 * either in-place edit or rename/copy edit.
2612                 */
2613                else if (DIFF_PAIR_RENAME(p)) {
2614                        /*
2615                         * A rename might have re-connected a broken
2616                         * pair up, causing the pathnames to be the
2617                         * same again. If so, that's not a rename at
2618                         * all, just a modification..
2619                         *
2620                         * Otherwise, see if this source was used for
2621                         * multiple renames, in which case we decrement
2622                         * the count, and call it a copy.
2623                         */
2624                        if (!strcmp(p->one->path, p->two->path))
2625                                p->status = DIFF_STATUS_MODIFIED;
2626                        else if (--p->one->rename_used > 0)
2627                                p->status = DIFF_STATUS_COPIED;
2628                        else
2629                                p->status = DIFF_STATUS_RENAMED;
2630                }
2631                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2632                         p->one->mode != p->two->mode ||
2633                         is_null_sha1(p->one->sha1))
2634                        p->status = DIFF_STATUS_MODIFIED;
2635                else {
2636                        /* This is a "no-change" entry and should not
2637                         * happen anymore, but prepare for broken callers.
2638                         */
2639                        error("feeding unmodified %s to diffcore",
2640                              p->one->path);
2641                        p->status = DIFF_STATUS_UNKNOWN;
2642                }
2643        }
2644        diff_debug_queue("resolve-rename-copy done", q);
2645}
2646
2647static int check_pair_status(struct diff_filepair *p)
2648{
2649        switch (p->status) {
2650        case DIFF_STATUS_UNKNOWN:
2651                return 0;
2652        case 0:
2653                die("internal error in diff-resolve-rename-copy");
2654        default:
2655                return 1;
2656        }
2657}
2658
2659static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2660{
2661        int fmt = opt->output_format;
2662
2663        if (fmt & DIFF_FORMAT_CHECKDIFF)
2664                diff_flush_checkdiff(p, opt);
2665        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2666                diff_flush_raw(p, opt);
2667        else if (fmt & DIFF_FORMAT_NAME)
2668                write_name_quoted(p->two->path, stdout, opt->line_termination);
2669}
2670
2671static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2672{
2673        if (fs->mode)
2674                printf(" %s mode %06o ", newdelete, fs->mode);
2675        else
2676                printf(" %s ", newdelete);
2677        write_name_quoted(fs->path, stdout, '\n');
2678}
2679
2680
2681static void show_mode_change(struct diff_filepair *p, int show_name)
2682{
2683        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2684                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2685                        show_name ? ' ' : '\n');
2686                if (show_name) {
2687                        write_name_quoted(p->two->path, stdout, '\n');
2688                }
2689        }
2690}
2691
2692static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2693{
2694        char *names = pprint_rename(p->one->path, p->two->path);
2695
2696        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2697        free(names);
2698        show_mode_change(p, 0);
2699}
2700
2701static void diff_summary(struct diff_filepair *p)
2702{
2703        switch(p->status) {
2704        case DIFF_STATUS_DELETED:
2705                show_file_mode_name("delete", p->one);
2706                break;
2707        case DIFF_STATUS_ADDED:
2708                show_file_mode_name("create", p->two);
2709                break;
2710        case DIFF_STATUS_COPIED:
2711                show_rename_copy("copy", p);
2712                break;
2713        case DIFF_STATUS_RENAMED:
2714                show_rename_copy("rename", p);
2715                break;
2716        default:
2717                if (p->score) {
2718                        puts(" rewrite ");
2719                        write_name_quoted(p->two->path, stdout, ' ');
2720                        printf("(%d%%)\n", similarity_index(p));
2721                }
2722                show_mode_change(p, !p->score);
2723                break;
2724        }
2725}
2726
2727struct patch_id_t {
2728        struct xdiff_emit_state xm;
2729        SHA_CTX *ctx;
2730        int patchlen;
2731};
2732
2733static int remove_space(char *line, int len)
2734{
2735        int i;
2736        char *dst = line;
2737        unsigned char c;
2738
2739        for (i = 0; i < len; i++)
2740                if (!isspace((c = line[i])))
2741                        *dst++ = c;
2742
2743        return dst - line;
2744}
2745
2746static void patch_id_consume(void *priv, char *line, unsigned long len)
2747{
2748        struct patch_id_t *data = priv;
2749        int new_len;
2750
2751        /* Ignore line numbers when computing the SHA1 of the patch */
2752        if (!prefixcmp(line, "@@ -"))
2753                return;
2754
2755        new_len = remove_space(line, len);
2756
2757        SHA1_Update(data->ctx, line, new_len);
2758        data->patchlen += new_len;
2759}
2760
2761/* returns 0 upon success, and writes result into sha1 */
2762static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2763{
2764        struct diff_queue_struct *q = &diff_queued_diff;
2765        int i;
2766        SHA_CTX ctx;
2767        struct patch_id_t data;
2768        char buffer[PATH_MAX * 4 + 20];
2769
2770        SHA1_Init(&ctx);
2771        memset(&data, 0, sizeof(struct patch_id_t));
2772        data.ctx = &ctx;
2773        data.xm.consume = patch_id_consume;
2774
2775        for (i = 0; i < q->nr; i++) {
2776                xpparam_t xpp;
2777                xdemitconf_t xecfg;
2778                xdemitcb_t ecb;
2779                mmfile_t mf1, mf2;
2780                struct diff_filepair *p = q->queue[i];
2781                int len1, len2;
2782
2783                memset(&xecfg, 0, sizeof(xecfg));
2784                if (p->status == 0)
2785                        return error("internal diff status error");
2786                if (p->status == DIFF_STATUS_UNKNOWN)
2787                        continue;
2788                if (diff_unmodified_pair(p))
2789                        continue;
2790                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2791                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2792                        continue;
2793                if (DIFF_PAIR_UNMERGED(p))
2794                        continue;
2795
2796                diff_fill_sha1_info(p->one);
2797                diff_fill_sha1_info(p->two);
2798                if (fill_mmfile(&mf1, p->one) < 0 ||
2799                                fill_mmfile(&mf2, p->two) < 0)
2800                        return error("unable to read files to diff");
2801
2802                len1 = remove_space(p->one->path, strlen(p->one->path));
2803                len2 = remove_space(p->two->path, strlen(p->two->path));
2804                if (p->one->mode == 0)
2805                        len1 = snprintf(buffer, sizeof(buffer),
2806                                        "diff--gita/%.*sb/%.*s"
2807                                        "newfilemode%06o"
2808                                        "---/dev/null"
2809                                        "+++b/%.*s",
2810                                        len1, p->one->path,
2811                                        len2, p->two->path,
2812                                        p->two->mode,
2813                                        len2, p->two->path);
2814                else if (p->two->mode == 0)
2815                        len1 = snprintf(buffer, sizeof(buffer),
2816                                        "diff--gita/%.*sb/%.*s"
2817                                        "deletedfilemode%06o"
2818                                        "---a/%.*s"
2819                                        "+++/dev/null",
2820                                        len1, p->one->path,
2821                                        len2, p->two->path,
2822                                        p->one->mode,
2823                                        len1, p->one->path);
2824                else
2825                        len1 = snprintf(buffer, sizeof(buffer),
2826                                        "diff--gita/%.*sb/%.*s"
2827                                        "---a/%.*s"
2828                                        "+++b/%.*s",
2829                                        len1, p->one->path,
2830                                        len2, p->two->path,
2831                                        len1, p->one->path,
2832                                        len2, p->two->path);
2833                SHA1_Update(&ctx, buffer, len1);
2834
2835                xpp.flags = XDF_NEED_MINIMAL;
2836                xecfg.ctxlen = 3;
2837                xecfg.flags = XDL_EMIT_FUNCNAMES;
2838                ecb.outf = xdiff_outf;
2839                ecb.priv = &data;
2840                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2841        }
2842
2843        SHA1_Final(sha1, &ctx);
2844        return 0;
2845}
2846
2847int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2848{
2849        struct diff_queue_struct *q = &diff_queued_diff;
2850        int i;
2851        int result = diff_get_patch_id(options, sha1);
2852
2853        for (i = 0; i < q->nr; i++)
2854                diff_free_filepair(q->queue[i]);
2855
2856        free(q->queue);
2857        q->queue = NULL;
2858        q->nr = q->alloc = 0;
2859
2860        return result;
2861}
2862
2863static int is_summary_empty(const struct diff_queue_struct *q)
2864{
2865        int i;
2866
2867        for (i = 0; i < q->nr; i++) {
2868                const struct diff_filepair *p = q->queue[i];
2869
2870                switch (p->status) {
2871                case DIFF_STATUS_DELETED:
2872                case DIFF_STATUS_ADDED:
2873                case DIFF_STATUS_COPIED:
2874                case DIFF_STATUS_RENAMED:
2875                        return 0;
2876                default:
2877                        if (p->score)
2878                                return 0;
2879                        if (p->one->mode && p->two->mode &&
2880                            p->one->mode != p->two->mode)
2881                                return 0;
2882                        break;
2883                }
2884        }
2885        return 1;
2886}
2887
2888void diff_flush(struct diff_options *options)
2889{
2890        struct diff_queue_struct *q = &diff_queued_diff;
2891        int i, output_format = options->output_format;
2892        int separator = 0;
2893
2894        /*
2895         * Order: raw, stat, summary, patch
2896         * or:    name/name-status/checkdiff (other bits clear)
2897         */
2898        if (!q->nr)
2899                goto free_queue;
2900
2901        if (output_format & (DIFF_FORMAT_RAW |
2902                             DIFF_FORMAT_NAME |
2903                             DIFF_FORMAT_NAME_STATUS |
2904                             DIFF_FORMAT_CHECKDIFF)) {
2905                for (i = 0; i < q->nr; i++) {
2906                        struct diff_filepair *p = q->queue[i];
2907                        if (check_pair_status(p))
2908                                flush_one_pair(p, options);
2909                }
2910                separator++;
2911        }
2912
2913        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2914                struct diffstat_t diffstat;
2915
2916                memset(&diffstat, 0, sizeof(struct diffstat_t));
2917                diffstat.xm.consume = diffstat_consume;
2918                for (i = 0; i < q->nr; i++) {
2919                        struct diff_filepair *p = q->queue[i];
2920                        if (check_pair_status(p))
2921                                diff_flush_stat(p, options, &diffstat);
2922                }
2923                if (output_format & DIFF_FORMAT_NUMSTAT)
2924                        show_numstat(&diffstat, options);
2925                if (output_format & DIFF_FORMAT_DIFFSTAT)
2926                        show_stats(&diffstat, options);
2927                else if (output_format & DIFF_FORMAT_SHORTSTAT)
2928                        show_shortstats(&diffstat);
2929                separator++;
2930        }
2931
2932        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2933                for (i = 0; i < q->nr; i++)
2934                        diff_summary(q->queue[i]);
2935                separator++;
2936        }
2937
2938        if (output_format & DIFF_FORMAT_PATCH) {
2939                if (separator) {
2940                        if (options->stat_sep) {
2941                                /* attach patch instead of inline */
2942                                fputs(options->stat_sep, stdout);
2943                        } else {
2944                                putchar(options->line_termination);
2945                        }
2946                }
2947
2948                for (i = 0; i < q->nr; i++) {
2949                        struct diff_filepair *p = q->queue[i];
2950                        if (check_pair_status(p))
2951                                diff_flush_patch(p, options);
2952                }
2953        }
2954
2955        if (output_format & DIFF_FORMAT_CALLBACK)
2956                options->format_callback(q, options, options->format_callback_data);
2957
2958        for (i = 0; i < q->nr; i++)
2959                diff_free_filepair(q->queue[i]);
2960free_queue:
2961        free(q->queue);
2962        q->queue = NULL;
2963        q->nr = q->alloc = 0;
2964}
2965
2966static void diffcore_apply_filter(const char *filter)
2967{
2968        int i;
2969        struct diff_queue_struct *q = &diff_queued_diff;
2970        struct diff_queue_struct outq;
2971        outq.queue = NULL;
2972        outq.nr = outq.alloc = 0;
2973
2974        if (!filter)
2975                return;
2976
2977        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2978                int found;
2979                for (i = found = 0; !found && i < q->nr; i++) {
2980                        struct diff_filepair *p = q->queue[i];
2981                        if (((p->status == DIFF_STATUS_MODIFIED) &&
2982                             ((p->score &&
2983                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2984                              (!p->score &&
2985                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2986                            ((p->status != DIFF_STATUS_MODIFIED) &&
2987                             strchr(filter, p->status)))
2988                                found++;
2989                }
2990                if (found)
2991                        return;
2992
2993                /* otherwise we will clear the whole queue
2994                 * by copying the empty outq at the end of this
2995                 * function, but first clear the current entries
2996                 * in the queue.
2997                 */
2998                for (i = 0; i < q->nr; i++)
2999                        diff_free_filepair(q->queue[i]);
3000        }
3001        else {
3002                /* Only the matching ones */
3003                for (i = 0; i < q->nr; i++) {
3004                        struct diff_filepair *p = q->queue[i];
3005
3006                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3007                             ((p->score &&
3008                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3009                              (!p->score &&
3010                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3011                            ((p->status != DIFF_STATUS_MODIFIED) &&
3012                             strchr(filter, p->status)))
3013                                diff_q(&outq, p);
3014                        else
3015                                diff_free_filepair(p);
3016                }
3017        }
3018        free(q->queue);
3019        *q = outq;
3020}
3021
3022/* Check whether two filespecs with the same mode and size are identical */
3023static int diff_filespec_is_identical(struct diff_filespec *one,
3024                                      struct diff_filespec *two)
3025{
3026        if (S_ISGITLINK(one->mode)) {
3027                diff_fill_sha1_info(one);
3028                diff_fill_sha1_info(two);
3029                return !hashcmp(one->sha1, two->sha1);
3030        }
3031        if (diff_populate_filespec(one, 0))
3032                return 0;
3033        if (diff_populate_filespec(two, 0))
3034                return 0;
3035        return !memcmp(one->data, two->data, one->size);
3036}
3037
3038static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3039{
3040        int i;
3041        struct diff_queue_struct *q = &diff_queued_diff;
3042        struct diff_queue_struct outq;
3043        outq.queue = NULL;
3044        outq.nr = outq.alloc = 0;
3045
3046        for (i = 0; i < q->nr; i++) {
3047                struct diff_filepair *p = q->queue[i];
3048
3049                /*
3050                 * 1. Entries that come from stat info dirtyness
3051                 *    always have both sides (iow, not create/delete),
3052                 *    one side of the object name is unknown, with
3053                 *    the same mode and size.  Keep the ones that
3054                 *    do not match these criteria.  They have real
3055                 *    differences.
3056                 *
3057                 * 2. At this point, the file is known to be modified,
3058                 *    with the same mode and size, and the object
3059                 *    name of one side is unknown.  Need to inspect
3060                 *    the identical contents.
3061                 */
3062                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3063                    !DIFF_FILE_VALID(p->two) ||
3064                    (p->one->sha1_valid && p->two->sha1_valid) ||
3065                    (p->one->mode != p->two->mode) ||
3066                    diff_populate_filespec(p->one, 1) ||
3067                    diff_populate_filespec(p->two, 1) ||
3068                    (p->one->size != p->two->size) ||
3069                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3070                        diff_q(&outq, p);
3071                else {
3072                        /*
3073                         * The caller can subtract 1 from skip_stat_unmatch
3074                         * to determine how many paths were dirty only
3075                         * due to stat info mismatch.
3076                         */
3077                        if (!diffopt->no_index)
3078                                diffopt->skip_stat_unmatch++;
3079                        diff_free_filepair(p);
3080                }
3081        }
3082        free(q->queue);
3083        *q = outq;
3084}
3085
3086void diffcore_std(struct diff_options *options)
3087{
3088        if (options->quiet)
3089                return;
3090
3091        if (options->skip_stat_unmatch && !options->find_copies_harder)
3092                diffcore_skip_stat_unmatch(options);
3093        if (options->break_opt != -1)
3094                diffcore_break(options->break_opt);
3095        if (options->detect_rename)
3096                diffcore_rename(options);
3097        if (options->break_opt != -1)
3098                diffcore_merge_broken();
3099        if (options->pickaxe)
3100                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3101        if (options->orderfile)
3102                diffcore_order(options->orderfile);
3103        diff_resolve_rename_copy();
3104        diffcore_apply_filter(options->filter);
3105
3106        options->has_changes = !!diff_queued_diff.nr;
3107}
3108
3109
3110void diff_addremove(struct diff_options *options,
3111                    int addremove, unsigned mode,
3112                    const unsigned char *sha1,
3113                    const char *base, const char *path)
3114{
3115        char concatpath[PATH_MAX];
3116        struct diff_filespec *one, *two;
3117
3118        /* This may look odd, but it is a preparation for
3119         * feeding "there are unchanged files which should
3120         * not produce diffs, but when you are doing copy
3121         * detection you would need them, so here they are"
3122         * entries to the diff-core.  They will be prefixed
3123         * with something like '=' or '*' (I haven't decided
3124         * which but should not make any difference).
3125         * Feeding the same new and old to diff_change()
3126         * also has the same effect.
3127         * Before the final output happens, they are pruned after
3128         * merged into rename/copy pairs as appropriate.
3129         */
3130        if (options->reverse_diff)
3131                addremove = (addremove == '+' ? '-' :
3132                             addremove == '-' ? '+' : addremove);
3133
3134        if (!path) path = "";
3135        sprintf(concatpath, "%s%s", base, path);
3136        one = alloc_filespec(concatpath);
3137        two = alloc_filespec(concatpath);
3138
3139        if (addremove != '+')
3140                fill_filespec(one, sha1, mode);
3141        if (addremove != '-')
3142                fill_filespec(two, sha1, mode);
3143
3144        diff_queue(&diff_queued_diff, one, two);
3145        options->has_changes = 1;
3146}
3147
3148void diff_change(struct diff_options *options,
3149                 unsigned old_mode, unsigned new_mode,
3150                 const unsigned char *old_sha1,
3151                 const unsigned char *new_sha1,
3152                 const char *base, const char *path)
3153{
3154        char concatpath[PATH_MAX];
3155        struct diff_filespec *one, *two;
3156
3157        if (options->reverse_diff) {
3158                unsigned tmp;
3159                const unsigned char *tmp_c;
3160                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3161                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3162        }
3163        if (!path) path = "";
3164        sprintf(concatpath, "%s%s", base, path);
3165        one = alloc_filespec(concatpath);
3166        two = alloc_filespec(concatpath);
3167        fill_filespec(one, old_sha1, old_mode);
3168        fill_filespec(two, new_sha1, new_mode);
3169
3170        diff_queue(&diff_queued_diff, one, two);
3171        options->has_changes = 1;
3172}
3173
3174void diff_unmerge(struct diff_options *options,
3175                  const char *path,
3176                  unsigned mode, const unsigned char *sha1)
3177{
3178        struct diff_filespec *one, *two;
3179        one = alloc_filespec(path);
3180        two = alloc_filespec(path);
3181        fill_filespec(one, sha1, mode);
3182        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3183}