diff.con commit Merge branch 'nd/worktree' into HEAD (7eedc1c)
   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
  13#ifdef NO_FAST_WORKING_DIRECTORY
  14#define FAST_WORKING_DIRECTORY 0
  15#else
  16#define FAST_WORKING_DIRECTORY 1
  17#endif
  18
  19static int diff_detect_rename_default;
  20static int diff_rename_limit_default = 100;
  21static int diff_use_color_default;
  22int diff_auto_refresh_index = 1;
  23
  24static char diff_colors[][COLOR_MAXLEN] = {
  25        "\033[m",       /* reset */
  26        "",             /* PLAIN (normal) */
  27        "\033[1m",      /* METAINFO (bold) */
  28        "\033[36m",     /* FRAGINFO (cyan) */
  29        "\033[31m",     /* OLD (red) */
  30        "\033[32m",     /* NEW (green) */
  31        "\033[33m",     /* COMMIT (yellow) */
  32        "\033[41m",     /* WHITESPACE (red background) */
  33};
  34
  35static int parse_diff_color_slot(const char *var, int ofs)
  36{
  37        if (!strcasecmp(var+ofs, "plain"))
  38                return DIFF_PLAIN;
  39        if (!strcasecmp(var+ofs, "meta"))
  40                return DIFF_METAINFO;
  41        if (!strcasecmp(var+ofs, "frag"))
  42                return DIFF_FRAGINFO;
  43        if (!strcasecmp(var+ofs, "old"))
  44                return DIFF_FILE_OLD;
  45        if (!strcasecmp(var+ofs, "new"))
  46                return DIFF_FILE_NEW;
  47        if (!strcasecmp(var+ofs, "commit"))
  48                return DIFF_COMMIT;
  49        if (!strcasecmp(var+ofs, "whitespace"))
  50                return DIFF_WHITESPACE;
  51        die("bad config variable '%s'", var);
  52}
  53
  54static struct ll_diff_driver {
  55        const char *name;
  56        struct ll_diff_driver *next;
  57        char *cmd;
  58} *user_diff, **user_diff_tail;
  59
  60static void read_config_if_needed(void)
  61{
  62        if (!user_diff_tail) {
  63                user_diff_tail = &user_diff;
  64                git_config(git_diff_ui_config);
  65        }
  66}
  67
  68/*
  69 * Currently there is only "diff.<drivername>.command" variable;
  70 * because there are "diff.color.<slot>" variables, we are parsing
  71 * this in a bit convoluted way to allow low level diff driver
  72 * called "color".
  73 */
  74static int parse_lldiff_command(const char *var, const char *ep, const char *value)
  75{
  76        const char *name;
  77        int namelen;
  78        struct ll_diff_driver *drv;
  79
  80        name = var + 5;
  81        namelen = ep - name;
  82        for (drv = user_diff; drv; drv = drv->next)
  83                if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
  84                        break;
  85        if (!drv) {
  86                drv = xcalloc(1, sizeof(struct ll_diff_driver));
  87                drv->name = xmemdupz(name, namelen);
  88                if (!user_diff_tail)
  89                        user_diff_tail = &user_diff;
  90                *user_diff_tail = drv;
  91                user_diff_tail = &(drv->next);
  92        }
  93
  94        if (!value)
  95                return error("%s: lacks value", var);
  96        drv->cmd = strdup(value);
  97        return 0;
  98}
  99
 100/*
 101 * 'diff.<what>.funcname' attribute can be specified in the configuration
 102 * to define a customized regexp to find the beginning of a function to
 103 * be used for hunk header lines of "diff -p" style output.
 104 */
 105static struct funcname_pattern {
 106        char *name;
 107        char *pattern;
 108        struct funcname_pattern *next;
 109} *funcname_pattern_list;
 110
 111static int parse_funcname_pattern(const char *var, const char *ep, const char *value)
 112{
 113        const char *name;
 114        int namelen;
 115        struct funcname_pattern *pp;
 116
 117        name = var + 5; /* "diff." */
 118        namelen = ep - name;
 119
 120        for (pp = funcname_pattern_list; pp; pp = pp->next)
 121                if (!strncmp(pp->name, name, namelen) && !pp->name[namelen])
 122                        break;
 123        if (!pp) {
 124                pp = xcalloc(1, sizeof(*pp));
 125                pp->name = xmemdupz(name, namelen);
 126                pp->next = funcname_pattern_list;
 127                funcname_pattern_list = pp;
 128        }
 129        if (pp->pattern)
 130                free(pp->pattern);
 131        pp->pattern = xstrdup(value);
 132        return 0;
 133}
 134
 135/*
 136 * These are to give UI layer defaults.
 137 * The core-level commands such as git-diff-files should
 138 * never be affected by the setting of diff.renames
 139 * the user happens to have in the configuration file.
 140 */
 141int git_diff_ui_config(const char *var, const char *value)
 142{
 143        if (!strcmp(var, "diff.renamelimit")) {
 144                diff_rename_limit_default = git_config_int(var, value);
 145                return 0;
 146        }
 147        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 148                diff_use_color_default = git_config_colorbool(var, value);
 149                return 0;
 150        }
 151        if (!strcmp(var, "diff.renames")) {
 152                if (!value)
 153                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 154                else if (!strcasecmp(value, "copies") ||
 155                         !strcasecmp(value, "copy"))
 156                        diff_detect_rename_default = DIFF_DETECT_COPY;
 157                else if (git_config_bool(var,value))
 158                        diff_detect_rename_default = DIFF_DETECT_RENAME;
 159                return 0;
 160        }
 161        if (!strcmp(var, "diff.autorefreshindex")) {
 162                diff_auto_refresh_index = git_config_bool(var, value);
 163                return 0;
 164        }
 165        if (!prefixcmp(var, "diff.")) {
 166                const char *ep = strrchr(var, '.');
 167
 168                if (ep != var + 4) {
 169                        if (!strcmp(ep, ".command"))
 170                                return parse_lldiff_command(var, ep, value);
 171                        if (!strcmp(ep, ".funcname"))
 172                                return parse_funcname_pattern(var, ep, value);
 173                }
 174        }
 175        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 176                int slot = parse_diff_color_slot(var, 11);
 177                color_parse(value, var, diff_colors[slot]);
 178                return 0;
 179        }
 180
 181        return git_default_config(var, value);
 182}
 183
 184static char *quote_two(const char *one, const char *two)
 185{
 186        int need_one = quote_c_style(one, NULL, NULL, 1);
 187        int need_two = quote_c_style(two, NULL, NULL, 1);
 188        struct strbuf res;
 189
 190        strbuf_init(&res, 0);
 191        if (need_one + need_two) {
 192                strbuf_addch(&res, '"');
 193                quote_c_style(one, &res, NULL, 1);
 194                quote_c_style(two, &res, NULL, 1);
 195                strbuf_addch(&res, '"');
 196        } else {
 197                strbuf_addstr(&res, one);
 198                strbuf_addstr(&res, two);
 199        }
 200        return strbuf_detach(&res, NULL);
 201}
 202
 203static const char *external_diff(void)
 204{
 205        static const char *external_diff_cmd = NULL;
 206        static int done_preparing = 0;
 207
 208        if (done_preparing)
 209                return external_diff_cmd;
 210        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 211        done_preparing = 1;
 212        return external_diff_cmd;
 213}
 214
 215static struct diff_tempfile {
 216        const char *name; /* filename external diff should read from */
 217        char hex[41];
 218        char mode[10];
 219        char tmp_path[PATH_MAX];
 220} diff_temp[2];
 221
 222static int count_lines(const char *data, int size)
 223{
 224        int count, ch, completely_empty = 1, nl_just_seen = 0;
 225        count = 0;
 226        while (0 < size--) {
 227                ch = *data++;
 228                if (ch == '\n') {
 229                        count++;
 230                        nl_just_seen = 1;
 231                        completely_empty = 0;
 232                }
 233                else {
 234                        nl_just_seen = 0;
 235                        completely_empty = 0;
 236                }
 237        }
 238        if (completely_empty)
 239                return 0;
 240        if (!nl_just_seen)
 241                count++; /* no trailing newline */
 242        return count;
 243}
 244
 245static void print_line_count(int count)
 246{
 247        switch (count) {
 248        case 0:
 249                printf("0,0");
 250                break;
 251        case 1:
 252                printf("1");
 253                break;
 254        default:
 255                printf("1,%d", count);
 256                break;
 257        }
 258}
 259
 260static void copy_file(int prefix, const char *data, int size,
 261                const char *set, const char *reset)
 262{
 263        int ch, nl_just_seen = 1;
 264        while (0 < size--) {
 265                ch = *data++;
 266                if (nl_just_seen) {
 267                        fputs(set, stdout);
 268                        putchar(prefix);
 269                }
 270                if (ch == '\n') {
 271                        nl_just_seen = 1;
 272                        fputs(reset, stdout);
 273                } else
 274                        nl_just_seen = 0;
 275                putchar(ch);
 276        }
 277        if (!nl_just_seen)
 278                printf("%s\n\\ No newline at end of file\n", reset);
 279}
 280
 281static void emit_rewrite_diff(const char *name_a,
 282                              const char *name_b,
 283                              struct diff_filespec *one,
 284                              struct diff_filespec *two,
 285                              int color_diff)
 286{
 287        int lc_a, lc_b;
 288        const char *name_a_tab, *name_b_tab;
 289        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 290        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 291        const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
 292        const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
 293        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 294
 295        name_a += (*name_a == '/');
 296        name_b += (*name_b == '/');
 297        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 298        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 299
 300        diff_populate_filespec(one, 0);
 301        diff_populate_filespec(two, 0);
 302        lc_a = count_lines(one->data, one->size);
 303        lc_b = count_lines(two->data, two->size);
 304        printf("%s--- a/%s%s%s\n%s+++ b/%s%s%s\n%s@@ -",
 305               metainfo, name_a, name_a_tab, reset,
 306               metainfo, name_b, name_b_tab, reset, fraginfo);
 307        print_line_count(lc_a);
 308        printf(" +");
 309        print_line_count(lc_b);
 310        printf(" @@%s\n", reset);
 311        if (lc_a)
 312                copy_file('-', one->data, one->size, old, reset);
 313        if (lc_b)
 314                copy_file('+', two->data, two->size, new, reset);
 315}
 316
 317static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 318{
 319        if (!DIFF_FILE_VALID(one)) {
 320                mf->ptr = (char *)""; /* does not matter */
 321                mf->size = 0;
 322                return 0;
 323        }
 324        else if (diff_populate_filespec(one, 0))
 325                return -1;
 326        mf->ptr = one->data;
 327        mf->size = one->size;
 328        return 0;
 329}
 330
 331struct diff_words_buffer {
 332        mmfile_t text;
 333        long alloc;
 334        long current; /* output pointer */
 335        int suppressed_newline;
 336};
 337
 338static void diff_words_append(char *line, unsigned long len,
 339                struct diff_words_buffer *buffer)
 340{
 341        if (buffer->text.size + len > buffer->alloc) {
 342                buffer->alloc = (buffer->text.size + len) * 3 / 2;
 343                buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
 344        }
 345        line++;
 346        len--;
 347        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 348        buffer->text.size += len;
 349}
 350
 351struct diff_words_data {
 352        struct xdiff_emit_state xm;
 353        struct diff_words_buffer minus, plus;
 354};
 355
 356static void print_word(struct diff_words_buffer *buffer, int len, int color,
 357                int suppress_newline)
 358{
 359        const char *ptr;
 360        int eol = 0;
 361
 362        if (len == 0)
 363                return;
 364
 365        ptr  = buffer->text.ptr + buffer->current;
 366        buffer->current += len;
 367
 368        if (ptr[len - 1] == '\n') {
 369                eol = 1;
 370                len--;
 371        }
 372
 373        fputs(diff_get_color(1, color), stdout);
 374        fwrite(ptr, len, 1, stdout);
 375        fputs(diff_get_color(1, DIFF_RESET), stdout);
 376
 377        if (eol) {
 378                if (suppress_newline)
 379                        buffer->suppressed_newline = 1;
 380                else
 381                        putchar('\n');
 382        }
 383}
 384
 385static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 386{
 387        struct diff_words_data *diff_words = priv;
 388
 389        if (diff_words->minus.suppressed_newline) {
 390                if (line[0] != '+')
 391                        putchar('\n');
 392                diff_words->minus.suppressed_newline = 0;
 393        }
 394
 395        len--;
 396        switch (line[0]) {
 397                case '-':
 398                        print_word(&diff_words->minus, len, DIFF_FILE_OLD, 1);
 399                        break;
 400                case '+':
 401                        print_word(&diff_words->plus, len, DIFF_FILE_NEW, 0);
 402                        break;
 403                case ' ':
 404                        print_word(&diff_words->plus, len, DIFF_PLAIN, 0);
 405                        diff_words->minus.current += len;
 406                        break;
 407        }
 408}
 409
 410/* this executes the word diff on the accumulated buffers */
 411static void diff_words_show(struct diff_words_data *diff_words)
 412{
 413        xpparam_t xpp;
 414        xdemitconf_t xecfg;
 415        xdemitcb_t ecb;
 416        mmfile_t minus, plus;
 417        int i;
 418
 419        memset(&xecfg, 0, sizeof(xecfg));
 420        minus.size = diff_words->minus.text.size;
 421        minus.ptr = xmalloc(minus.size);
 422        memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
 423        for (i = 0; i < minus.size; i++)
 424                if (isspace(minus.ptr[i]))
 425                        minus.ptr[i] = '\n';
 426        diff_words->minus.current = 0;
 427
 428        plus.size = diff_words->plus.text.size;
 429        plus.ptr = xmalloc(plus.size);
 430        memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
 431        for (i = 0; i < plus.size; i++)
 432                if (isspace(plus.ptr[i]))
 433                        plus.ptr[i] = '\n';
 434        diff_words->plus.current = 0;
 435
 436        xpp.flags = XDF_NEED_MINIMAL;
 437        xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
 438        ecb.outf = xdiff_outf;
 439        ecb.priv = diff_words;
 440        diff_words->xm.consume = fn_out_diff_words_aux;
 441        xdl_diff(&minus, &plus, &xpp, &xecfg, &ecb);
 442
 443        free(minus.ptr);
 444        free(plus.ptr);
 445        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 446
 447        if (diff_words->minus.suppressed_newline) {
 448                putchar('\n');
 449                diff_words->minus.suppressed_newline = 0;
 450        }
 451}
 452
 453struct emit_callback {
 454        struct xdiff_emit_state xm;
 455        int nparents, color_diff;
 456        const char **label_path;
 457        struct diff_words_data *diff_words;
 458        int *found_changesp;
 459};
 460
 461static void free_diff_words_data(struct emit_callback *ecbdata)
 462{
 463        if (ecbdata->diff_words) {
 464                /* flush buffers */
 465                if (ecbdata->diff_words->minus.text.size ||
 466                                ecbdata->diff_words->plus.text.size)
 467                        diff_words_show(ecbdata->diff_words);
 468
 469                if (ecbdata->diff_words->minus.text.ptr)
 470                        free (ecbdata->diff_words->minus.text.ptr);
 471                if (ecbdata->diff_words->plus.text.ptr)
 472                        free (ecbdata->diff_words->plus.text.ptr);
 473                free(ecbdata->diff_words);
 474                ecbdata->diff_words = NULL;
 475        }
 476}
 477
 478const char *diff_get_color(int diff_use_color, enum color_diff ix)
 479{
 480        if (diff_use_color)
 481                return diff_colors[ix];
 482        return "";
 483}
 484
 485static void emit_line(const char *set, const char *reset, const char *line, int len)
 486{
 487        if (len > 0 && line[len-1] == '\n')
 488                len--;
 489        fputs(set, stdout);
 490        fwrite(line, len, 1, stdout);
 491        puts(reset);
 492}
 493
 494static void emit_line_with_ws(int nparents,
 495                const char *set, const char *reset, const char *ws,
 496                const char *line, int len)
 497{
 498        int col0 = nparents;
 499        int last_tab_in_indent = -1;
 500        int last_space_in_indent = -1;
 501        int i;
 502        int tail = len;
 503        int need_highlight_leading_space = 0;
 504        /* The line is a newly added line.  Does it have funny leading
 505         * whitespaces?  In indent, SP should never precede a TAB.
 506         */
 507        for (i = col0; i < len; i++) {
 508                if (line[i] == '\t') {
 509                        last_tab_in_indent = i;
 510                        if (0 <= last_space_in_indent)
 511                                need_highlight_leading_space = 1;
 512                }
 513                else if (line[i] == ' ')
 514                        last_space_in_indent = i;
 515                else
 516                        break;
 517        }
 518        fputs(set, stdout);
 519        fwrite(line, col0, 1, stdout);
 520        fputs(reset, stdout);
 521        if (((i == len) || line[i] == '\n') && i != col0) {
 522                /* The whole line was indent */
 523                emit_line(ws, reset, line + col0, len - col0);
 524                return;
 525        }
 526        i = col0;
 527        if (need_highlight_leading_space) {
 528                while (i < last_tab_in_indent) {
 529                        if (line[i] == ' ') {
 530                                fputs(ws, stdout);
 531                                putchar(' ');
 532                                fputs(reset, stdout);
 533                        }
 534                        else
 535                                putchar(line[i]);
 536                        i++;
 537                }
 538        }
 539        tail = len - 1;
 540        if (line[tail] == '\n' && i < tail)
 541                tail--;
 542        while (i < tail) {
 543                if (!isspace(line[tail]))
 544                        break;
 545                tail--;
 546        }
 547        if ((i < tail && line[tail + 1] != '\n')) {
 548                /* This has whitespace between tail+1..len */
 549                fputs(set, stdout);
 550                fwrite(line + i, tail - i + 1, 1, stdout);
 551                fputs(reset, stdout);
 552                emit_line(ws, reset, line + tail + 1, len - tail - 1);
 553        }
 554        else
 555                emit_line(set, reset, line + i, len - i);
 556}
 557
 558static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
 559{
 560        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 561        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 562
 563        if (!*ws)
 564                emit_line(set, reset, line, len);
 565        else
 566                emit_line_with_ws(ecbdata->nparents, set, reset, ws,
 567                                line, len);
 568}
 569
 570static void fn_out_consume(void *priv, char *line, unsigned long len)
 571{
 572        int i;
 573        int color;
 574        struct emit_callback *ecbdata = priv;
 575        const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 576        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 577
 578        *(ecbdata->found_changesp) = 1;
 579
 580        if (ecbdata->label_path[0]) {
 581                const char *name_a_tab, *name_b_tab;
 582
 583                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 584                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 585
 586                printf("%s--- %s%s%s\n",
 587                       set, ecbdata->label_path[0], reset, name_a_tab);
 588                printf("%s+++ %s%s%s\n",
 589                       set, ecbdata->label_path[1], reset, name_b_tab);
 590                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 591        }
 592
 593        /* This is not really necessary for now because
 594         * this codepath only deals with two-way diffs.
 595         */
 596        for (i = 0; i < len && line[i] == '@'; i++)
 597                ;
 598        if (2 <= i && i < len && line[i] == ' ') {
 599                ecbdata->nparents = i - 1;
 600                emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 601                          reset, line, len);
 602                return;
 603        }
 604
 605        if (len < ecbdata->nparents) {
 606                set = reset;
 607                emit_line(reset, reset, line, len);
 608                return;
 609        }
 610
 611        color = DIFF_PLAIN;
 612        if (ecbdata->diff_words && ecbdata->nparents != 1)
 613                /* fall back to normal diff */
 614                free_diff_words_data(ecbdata);
 615        if (ecbdata->diff_words) {
 616                if (line[0] == '-') {
 617                        diff_words_append(line, len,
 618                                          &ecbdata->diff_words->minus);
 619                        return;
 620                } else if (line[0] == '+') {
 621                        diff_words_append(line, len,
 622                                          &ecbdata->diff_words->plus);
 623                        return;
 624                }
 625                if (ecbdata->diff_words->minus.text.size ||
 626                    ecbdata->diff_words->plus.text.size)
 627                        diff_words_show(ecbdata->diff_words);
 628                line++;
 629                len--;
 630                emit_line(set, reset, line, len);
 631                return;
 632        }
 633        for (i = 0; i < ecbdata->nparents && len; i++) {
 634                if (line[i] == '-')
 635                        color = DIFF_FILE_OLD;
 636                else if (line[i] == '+')
 637                        color = DIFF_FILE_NEW;
 638        }
 639
 640        if (color != DIFF_FILE_NEW) {
 641                emit_line(diff_get_color(ecbdata->color_diff, color),
 642                          reset, line, len);
 643                return;
 644        }
 645        emit_add_line(reset, ecbdata, line, len);
 646}
 647
 648static char *pprint_rename(const char *a, const char *b)
 649{
 650        const char *old = a;
 651        const char *new = b;
 652        struct strbuf name;
 653        int pfx_length, sfx_length;
 654        int len_a = strlen(a);
 655        int len_b = strlen(b);
 656        int a_midlen, b_midlen;
 657        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 658        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 659
 660        strbuf_init(&name, 0);
 661        if (qlen_a || qlen_b) {
 662                quote_c_style(a, &name, NULL, 0);
 663                strbuf_addstr(&name, " => ");
 664                quote_c_style(b, &name, NULL, 0);
 665                return strbuf_detach(&name, NULL);
 666        }
 667
 668        /* Find common prefix */
 669        pfx_length = 0;
 670        while (*old && *new && *old == *new) {
 671                if (*old == '/')
 672                        pfx_length = old - a + 1;
 673                old++;
 674                new++;
 675        }
 676
 677        /* Find common suffix */
 678        old = a + len_a;
 679        new = b + len_b;
 680        sfx_length = 0;
 681        while (a <= old && b <= new && *old == *new) {
 682                if (*old == '/')
 683                        sfx_length = len_a - (old - a);
 684                old--;
 685                new--;
 686        }
 687
 688        /*
 689         * pfx{mid-a => mid-b}sfx
 690         * {pfx-a => pfx-b}sfx
 691         * pfx{sfx-a => sfx-b}
 692         * name-a => name-b
 693         */
 694        a_midlen = len_a - pfx_length - sfx_length;
 695        b_midlen = len_b - pfx_length - sfx_length;
 696        if (a_midlen < 0)
 697                a_midlen = 0;
 698        if (b_midlen < 0)
 699                b_midlen = 0;
 700
 701        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 702        if (pfx_length + sfx_length) {
 703                strbuf_add(&name, a, pfx_length);
 704                strbuf_addch(&name, '{');
 705        }
 706        strbuf_add(&name, a + pfx_length, a_midlen);
 707        strbuf_addstr(&name, " => ");
 708        strbuf_add(&name, b + pfx_length, b_midlen);
 709        if (pfx_length + sfx_length) {
 710                strbuf_addch(&name, '}');
 711                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 712        }
 713        return strbuf_detach(&name, NULL);
 714}
 715
 716struct diffstat_t {
 717        struct xdiff_emit_state xm;
 718
 719        int nr;
 720        int alloc;
 721        struct diffstat_file {
 722                char *name;
 723                unsigned is_unmerged:1;
 724                unsigned is_binary:1;
 725                unsigned is_renamed:1;
 726                unsigned int added, deleted;
 727        } **files;
 728};
 729
 730static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 731                                          const char *name_a,
 732                                          const char *name_b)
 733{
 734        struct diffstat_file *x;
 735        x = xcalloc(sizeof (*x), 1);
 736        if (diffstat->nr == diffstat->alloc) {
 737                diffstat->alloc = alloc_nr(diffstat->alloc);
 738                diffstat->files = xrealloc(diffstat->files,
 739                                diffstat->alloc * sizeof(x));
 740        }
 741        diffstat->files[diffstat->nr++] = x;
 742        if (name_b) {
 743                x->name = pprint_rename(name_a, name_b);
 744                x->is_renamed = 1;
 745        }
 746        else
 747                x->name = xstrdup(name_a);
 748        return x;
 749}
 750
 751static void diffstat_consume(void *priv, char *line, unsigned long len)
 752{
 753        struct diffstat_t *diffstat = priv;
 754        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 755
 756        if (line[0] == '+')
 757                x->added++;
 758        else if (line[0] == '-')
 759                x->deleted++;
 760}
 761
 762const char mime_boundary_leader[] = "------------";
 763
 764static int scale_linear(int it, int width, int max_change)
 765{
 766        /*
 767         * make sure that at least one '-' is printed if there were deletions,
 768         * and likewise for '+'.
 769         */
 770        if (max_change < 2)
 771                return it;
 772        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 773}
 774
 775static void show_name(const char *prefix, const char *name, int len,
 776                      const char *reset, const char *set)
 777{
 778        printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
 779}
 780
 781static void show_graph(char ch, int cnt, const char *set, const char *reset)
 782{
 783        if (cnt <= 0)
 784                return;
 785        printf("%s", set);
 786        while (cnt--)
 787                putchar(ch);
 788        printf("%s", reset);
 789}
 790
 791static void show_stats(struct diffstat_t* data, struct diff_options *options)
 792{
 793        int i, len, add, del, total, adds = 0, dels = 0;
 794        int max_change = 0, max_len = 0;
 795        int total_files = data->nr;
 796        int width, name_width;
 797        const char *reset, *set, *add_c, *del_c;
 798
 799        if (data->nr == 0)
 800                return;
 801
 802        width = options->stat_width ? options->stat_width : 80;
 803        name_width = options->stat_name_width ? options->stat_name_width : 50;
 804
 805        /* Sanity: give at least 5 columns to the graph,
 806         * but leave at least 10 columns for the name.
 807         */
 808        if (width < name_width + 15) {
 809                if (name_width <= 25)
 810                        width = name_width + 15;
 811                else
 812                        name_width = width - 15;
 813        }
 814
 815        /* Find the longest filename and max number of changes */
 816        reset = diff_get_color(options->color_diff, DIFF_RESET);
 817        set = diff_get_color(options->color_diff, DIFF_PLAIN);
 818        add_c = diff_get_color(options->color_diff, DIFF_FILE_NEW);
 819        del_c = diff_get_color(options->color_diff, DIFF_FILE_OLD);
 820
 821        for (i = 0; i < data->nr; i++) {
 822                struct diffstat_file *file = data->files[i];
 823                int change = file->added + file->deleted;
 824
 825                if (!file->is_renamed) {  /* renames are already quoted by pprint_rename */
 826                        struct strbuf buf;
 827                        strbuf_init(&buf, 0);
 828                        if (quote_c_style(file->name, &buf, NULL, 0)) {
 829                                free(file->name);
 830                                file->name = strbuf_detach(&buf, NULL);
 831                        } else {
 832                                strbuf_release(&buf);
 833                        }
 834                }
 835
 836                len = strlen(file->name);
 837                if (max_len < len)
 838                        max_len = len;
 839
 840                if (file->is_binary || file->is_unmerged)
 841                        continue;
 842                if (max_change < change)
 843                        max_change = change;
 844        }
 845
 846        /* Compute the width of the graph part;
 847         * 10 is for one blank at the beginning of the line plus
 848         * " | count " between the name and the graph.
 849         *
 850         * From here on, name_width is the width of the name area,
 851         * and width is the width of the graph area.
 852         */
 853        name_width = (name_width < max_len) ? name_width : max_len;
 854        if (width < (name_width + 10) + max_change)
 855                width = width - (name_width + 10);
 856        else
 857                width = max_change;
 858
 859        for (i = 0; i < data->nr; i++) {
 860                const char *prefix = "";
 861                char *name = data->files[i]->name;
 862                int added = data->files[i]->added;
 863                int deleted = data->files[i]->deleted;
 864                int name_len;
 865
 866                /*
 867                 * "scale" the filename
 868                 */
 869                len = name_width;
 870                name_len = strlen(name);
 871                if (name_width < name_len) {
 872                        char *slash;
 873                        prefix = "...";
 874                        len -= 3;
 875                        name += name_len - len;
 876                        slash = strchr(name, '/');
 877                        if (slash)
 878                                name = slash;
 879                }
 880
 881                if (data->files[i]->is_binary) {
 882                        show_name(prefix, name, len, reset, set);
 883                        printf("  Bin ");
 884                        printf("%s%d%s", del_c, deleted, reset);
 885                        printf(" -> ");
 886                        printf("%s%d%s", add_c, added, reset);
 887                        printf(" bytes");
 888                        printf("\n");
 889                        goto free_diffstat_file;
 890                }
 891                else if (data->files[i]->is_unmerged) {
 892                        show_name(prefix, name, len, reset, set);
 893                        printf("  Unmerged\n");
 894                        goto free_diffstat_file;
 895                }
 896                else if (!data->files[i]->is_renamed &&
 897                         (added + deleted == 0)) {
 898                        total_files--;
 899                        goto free_diffstat_file;
 900                }
 901
 902                /*
 903                 * scale the add/delete
 904                 */
 905                add = added;
 906                del = deleted;
 907                total = add + del;
 908                adds += add;
 909                dels += del;
 910
 911                if (width <= max_change) {
 912                        add = scale_linear(add, width, max_change);
 913                        del = scale_linear(del, width, max_change);
 914                        total = add + del;
 915                }
 916                show_name(prefix, name, len, reset, set);
 917                printf("%5d ", added + deleted);
 918                show_graph('+', add, add_c, reset);
 919                show_graph('-', del, del_c, reset);
 920                putchar('\n');
 921        free_diffstat_file:
 922                free(data->files[i]->name);
 923                free(data->files[i]);
 924        }
 925        free(data->files);
 926        printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
 927               set, total_files, adds, dels, reset);
 928}
 929
 930static void show_shortstats(struct diffstat_t* data)
 931{
 932        int i, adds = 0, dels = 0, total_files = data->nr;
 933
 934        if (data->nr == 0)
 935                return;
 936
 937        for (i = 0; i < data->nr; i++) {
 938                if (!data->files[i]->is_binary &&
 939                    !data->files[i]->is_unmerged) {
 940                        int added = data->files[i]->added;
 941                        int deleted= data->files[i]->deleted;
 942                        if (!data->files[i]->is_renamed &&
 943                            (added + deleted == 0)) {
 944                                total_files--;
 945                        } else {
 946                                adds += added;
 947                                dels += deleted;
 948                        }
 949                }
 950                free(data->files[i]->name);
 951                free(data->files[i]);
 952        }
 953        free(data->files);
 954
 955        printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
 956               total_files, adds, dels);
 957}
 958
 959static void show_numstat(struct diffstat_t* data, struct diff_options *options)
 960{
 961        int i;
 962
 963        for (i = 0; i < data->nr; i++) {
 964                struct diffstat_file *file = data->files[i];
 965
 966                if (file->is_binary)
 967                        printf("-\t-\t");
 968                else
 969                        printf("%d\t%d\t", file->added, file->deleted);
 970                if (!file->is_renamed) {
 971                        write_name_quoted(file->name, stdout, options->line_termination);
 972                } else {
 973                        fputs(file->name, stdout);
 974                        putchar(options->line_termination);
 975                }
 976        }
 977}
 978
 979struct checkdiff_t {
 980        struct xdiff_emit_state xm;
 981        const char *filename;
 982        int lineno, color_diff;
 983};
 984
 985static void checkdiff_consume(void *priv, char *line, unsigned long len)
 986{
 987        struct checkdiff_t *data = priv;
 988        const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
 989        const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
 990        const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
 991
 992        if (line[0] == '+') {
 993                int i, spaces = 0, space_before_tab = 0, white_space_at_end = 0;
 994
 995                /* check space before tab */
 996                for (i = 1; i < len && (line[i] == ' ' || line[i] == '\t'); i++)
 997                        if (line[i] == ' ')
 998                                spaces++;
 999                if (line[i - 1] == '\t' && spaces)
1000                        space_before_tab = 1;
1001
1002                /* check white space at line end */
1003                if (line[len - 1] == '\n')
1004                        len--;
1005                if (isspace(line[len - 1]))
1006                        white_space_at_end = 1;
1007
1008                if (space_before_tab || white_space_at_end) {
1009                        printf("%s:%d: %s", data->filename, data->lineno, ws);
1010                        if (space_before_tab) {
1011                                printf("space before tab");
1012                                if (white_space_at_end)
1013                                        putchar(',');
1014                        }
1015                        if (white_space_at_end)
1016                                printf("white space at end");
1017                        printf(":%s ", reset);
1018                        emit_line_with_ws(1, set, reset, ws, line, len);
1019                }
1020
1021                data->lineno++;
1022        } else if (line[0] == ' ')
1023                data->lineno++;
1024        else if (line[0] == '@') {
1025                char *plus = strchr(line, '+');
1026                if (plus)
1027                        data->lineno = strtol(plus, NULL, 10);
1028                else
1029                        die("invalid diff");
1030        }
1031}
1032
1033static unsigned char *deflate_it(char *data,
1034                                 unsigned long size,
1035                                 unsigned long *result_size)
1036{
1037        int bound;
1038        unsigned char *deflated;
1039        z_stream stream;
1040
1041        memset(&stream, 0, sizeof(stream));
1042        deflateInit(&stream, zlib_compression_level);
1043        bound = deflateBound(&stream, size);
1044        deflated = xmalloc(bound);
1045        stream.next_out = deflated;
1046        stream.avail_out = bound;
1047
1048        stream.next_in = (unsigned char *)data;
1049        stream.avail_in = size;
1050        while (deflate(&stream, Z_FINISH) == Z_OK)
1051                ; /* nothing */
1052        deflateEnd(&stream);
1053        *result_size = stream.total_out;
1054        return deflated;
1055}
1056
1057static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
1058{
1059        void *cp;
1060        void *delta;
1061        void *deflated;
1062        void *data;
1063        unsigned long orig_size;
1064        unsigned long delta_size;
1065        unsigned long deflate_size;
1066        unsigned long data_size;
1067
1068        /* We could do deflated delta, or we could do just deflated two,
1069         * whichever is smaller.
1070         */
1071        delta = NULL;
1072        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1073        if (one->size && two->size) {
1074                delta = diff_delta(one->ptr, one->size,
1075                                   two->ptr, two->size,
1076                                   &delta_size, deflate_size);
1077                if (delta) {
1078                        void *to_free = delta;
1079                        orig_size = delta_size;
1080                        delta = deflate_it(delta, delta_size, &delta_size);
1081                        free(to_free);
1082                }
1083        }
1084
1085        if (delta && delta_size < deflate_size) {
1086                printf("delta %lu\n", orig_size);
1087                free(deflated);
1088                data = delta;
1089                data_size = delta_size;
1090        }
1091        else {
1092                printf("literal %lu\n", two->size);
1093                free(delta);
1094                data = deflated;
1095                data_size = deflate_size;
1096        }
1097
1098        /* emit data encoded in base85 */
1099        cp = data;
1100        while (data_size) {
1101                int bytes = (52 < data_size) ? 52 : data_size;
1102                char line[70];
1103                data_size -= bytes;
1104                if (bytes <= 26)
1105                        line[0] = bytes + 'A' - 1;
1106                else
1107                        line[0] = bytes - 26 + 'a' - 1;
1108                encode_85(line + 1, cp, bytes);
1109                cp = (char *) cp + bytes;
1110                puts(line);
1111        }
1112        printf("\n");
1113        free(data);
1114}
1115
1116static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1117{
1118        printf("GIT binary patch\n");
1119        emit_binary_diff_body(one, two);
1120        emit_binary_diff_body(two, one);
1121}
1122
1123static void setup_diff_attr_check(struct git_attr_check *check)
1124{
1125        static struct git_attr *attr_diff;
1126
1127        if (!attr_diff) {
1128                attr_diff = git_attr("diff", 4);
1129        }
1130        check[0].attr = attr_diff;
1131}
1132
1133static void diff_filespec_check_attr(struct diff_filespec *one)
1134{
1135        struct git_attr_check attr_diff_check;
1136        int check_from_data = 0;
1137
1138        if (one->checked_attr)
1139                return;
1140
1141        setup_diff_attr_check(&attr_diff_check);
1142        one->is_binary = 0;
1143        one->funcname_pattern_ident = NULL;
1144
1145        if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1146                const char *value;
1147
1148                /* binaryness */
1149                value = attr_diff_check.value;
1150                if (ATTR_TRUE(value))
1151                        ;
1152                else if (ATTR_FALSE(value))
1153                        one->is_binary = 1;
1154                else
1155                        check_from_data = 1;
1156
1157                /* funcname pattern ident */
1158                if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1159                        ;
1160                else
1161                        one->funcname_pattern_ident = value;
1162        }
1163
1164        if (check_from_data) {
1165                if (!one->data && DIFF_FILE_VALID(one))
1166                        diff_populate_filespec(one, 0);
1167
1168                if (one->data)
1169                        one->is_binary = buffer_is_binary(one->data, one->size);
1170        }
1171}
1172
1173int diff_filespec_is_binary(struct diff_filespec *one)
1174{
1175        diff_filespec_check_attr(one);
1176        return one->is_binary;
1177}
1178
1179static const char *funcname_pattern(const char *ident)
1180{
1181        struct funcname_pattern *pp;
1182
1183        read_config_if_needed();
1184        for (pp = funcname_pattern_list; pp; pp = pp->next)
1185                if (!strcmp(ident, pp->name))
1186                        return pp->pattern;
1187        return NULL;
1188}
1189
1190static struct builtin_funcname_pattern {
1191        const char *name;
1192        const char *pattern;
1193} builtin_funcname_pattern[] = {
1194        { "java", "!^[  ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1195                        "new\\|return\\|switch\\|throw\\|while\\)\n"
1196                        "^[     ]*\\(\\([       ]*"
1197                        "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1198                        "[      ]*([^;]*$\\)" },
1199        { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1200};
1201
1202static const char *diff_funcname_pattern(struct diff_filespec *one)
1203{
1204        const char *ident, *pattern;
1205        int i;
1206
1207        diff_filespec_check_attr(one);
1208        ident = one->funcname_pattern_ident;
1209
1210        if (!ident)
1211                /*
1212                 * If the config file has "funcname.default" defined, that
1213                 * regexp is used; otherwise NULL is returned and xemit uses
1214                 * the built-in default.
1215                 */
1216                return funcname_pattern("default");
1217
1218        /* Look up custom "funcname.$ident" regexp from config. */
1219        pattern = funcname_pattern(ident);
1220        if (pattern)
1221                return pattern;
1222
1223        /*
1224         * And define built-in fallback patterns here.  Note that
1225         * these can be overriden by the user's config settings.
1226         */
1227        for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1228                if (!strcmp(ident, builtin_funcname_pattern[i].name))
1229                        return builtin_funcname_pattern[i].pattern;
1230
1231        return NULL;
1232}
1233
1234static void builtin_diff(const char *name_a,
1235                         const char *name_b,
1236                         struct diff_filespec *one,
1237                         struct diff_filespec *two,
1238                         const char *xfrm_msg,
1239                         struct diff_options *o,
1240                         int complete_rewrite)
1241{
1242        mmfile_t mf1, mf2;
1243        const char *lbl[2];
1244        char *a_one, *b_two;
1245        const char *set = diff_get_color(o->color_diff, DIFF_METAINFO);
1246        const char *reset = diff_get_color(o->color_diff, DIFF_RESET);
1247
1248        a_one = quote_two("a/", name_a + (*name_a == '/'));
1249        b_two = quote_two("b/", name_b + (*name_b == '/'));
1250        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1251        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1252        printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1253        if (lbl[0][0] == '/') {
1254                /* /dev/null */
1255                printf("%snew file mode %06o%s\n", set, two->mode, reset);
1256                if (xfrm_msg && xfrm_msg[0])
1257                        printf("%s%s%s\n", set, xfrm_msg, reset);
1258        }
1259        else if (lbl[1][0] == '/') {
1260                printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1261                if (xfrm_msg && xfrm_msg[0])
1262                        printf("%s%s%s\n", set, xfrm_msg, reset);
1263        }
1264        else {
1265                if (one->mode != two->mode) {
1266                        printf("%sold mode %06o%s\n", set, one->mode, reset);
1267                        printf("%snew mode %06o%s\n", set, two->mode, reset);
1268                }
1269                if (xfrm_msg && xfrm_msg[0])
1270                        printf("%s%s%s\n", set, xfrm_msg, reset);
1271                /*
1272                 * we do not run diff between different kind
1273                 * of objects.
1274                 */
1275                if ((one->mode ^ two->mode) & S_IFMT)
1276                        goto free_ab_and_return;
1277                if (complete_rewrite) {
1278                        emit_rewrite_diff(name_a, name_b, one, two,
1279                                        o->color_diff);
1280                        o->found_changes = 1;
1281                        goto free_ab_and_return;
1282                }
1283        }
1284
1285        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1286                die("unable to read files to diff");
1287
1288        if (!o->text &&
1289            (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1290                /* Quite common confusing case */
1291                if (mf1.size == mf2.size &&
1292                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1293                        goto free_ab_and_return;
1294                if (o->binary)
1295                        emit_binary_diff(&mf1, &mf2);
1296                else
1297                        printf("Binary files %s and %s differ\n",
1298                               lbl[0], lbl[1]);
1299                o->found_changes = 1;
1300        }
1301        else {
1302                /* Crazy xdl interfaces.. */
1303                const char *diffopts = getenv("GIT_DIFF_OPTS");
1304                xpparam_t xpp;
1305                xdemitconf_t xecfg;
1306                xdemitcb_t ecb;
1307                struct emit_callback ecbdata;
1308                const char *funcname_pattern;
1309
1310                funcname_pattern = diff_funcname_pattern(one);
1311                if (!funcname_pattern)
1312                        funcname_pattern = diff_funcname_pattern(two);
1313
1314                memset(&xecfg, 0, sizeof(xecfg));
1315                memset(&ecbdata, 0, sizeof(ecbdata));
1316                ecbdata.label_path = lbl;
1317                ecbdata.color_diff = o->color_diff;
1318                ecbdata.found_changesp = &o->found_changes;
1319                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1320                xecfg.ctxlen = o->context;
1321                xecfg.flags = XDL_EMIT_FUNCNAMES;
1322                if (funcname_pattern)
1323                        xdiff_set_find_func(&xecfg, funcname_pattern);
1324                if (!diffopts)
1325                        ;
1326                else if (!prefixcmp(diffopts, "--unified="))
1327                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1328                else if (!prefixcmp(diffopts, "-u"))
1329                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1330                ecb.outf = xdiff_outf;
1331                ecb.priv = &ecbdata;
1332                ecbdata.xm.consume = fn_out_consume;
1333                if (o->color_diff_words)
1334                        ecbdata.diff_words =
1335                                xcalloc(1, sizeof(struct diff_words_data));
1336                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1337                if (o->color_diff_words)
1338                        free_diff_words_data(&ecbdata);
1339        }
1340
1341 free_ab_and_return:
1342        diff_free_filespec_data(one);
1343        diff_free_filespec_data(two);
1344        free(a_one);
1345        free(b_two);
1346        return;
1347}
1348
1349static void builtin_diffstat(const char *name_a, const char *name_b,
1350                             struct diff_filespec *one,
1351                             struct diff_filespec *two,
1352                             struct diffstat_t *diffstat,
1353                             struct diff_options *o,
1354                             int complete_rewrite)
1355{
1356        mmfile_t mf1, mf2;
1357        struct diffstat_file *data;
1358
1359        data = diffstat_add(diffstat, name_a, name_b);
1360
1361        if (!one || !two) {
1362                data->is_unmerged = 1;
1363                return;
1364        }
1365        if (complete_rewrite) {
1366                diff_populate_filespec(one, 0);
1367                diff_populate_filespec(two, 0);
1368                data->deleted = count_lines(one->data, one->size);
1369                data->added = count_lines(two->data, two->size);
1370                goto free_and_return;
1371        }
1372        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1373                die("unable to read files to diff");
1374
1375        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1376                data->is_binary = 1;
1377                data->added = mf2.size;
1378                data->deleted = mf1.size;
1379        } else {
1380                /* Crazy xdl interfaces.. */
1381                xpparam_t xpp;
1382                xdemitconf_t xecfg;
1383                xdemitcb_t ecb;
1384
1385                memset(&xecfg, 0, sizeof(xecfg));
1386                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1387                ecb.outf = xdiff_outf;
1388                ecb.priv = diffstat;
1389                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1390        }
1391
1392 free_and_return:
1393        diff_free_filespec_data(one);
1394        diff_free_filespec_data(two);
1395}
1396
1397static void builtin_checkdiff(const char *name_a, const char *name_b,
1398                             struct diff_filespec *one,
1399                             struct diff_filespec *two, struct diff_options *o)
1400{
1401        mmfile_t mf1, mf2;
1402        struct checkdiff_t data;
1403
1404        if (!two)
1405                return;
1406
1407        memset(&data, 0, sizeof(data));
1408        data.xm.consume = checkdiff_consume;
1409        data.filename = name_b ? name_b : name_a;
1410        data.lineno = 0;
1411        data.color_diff = o->color_diff;
1412
1413        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1414                die("unable to read files to diff");
1415
1416        if (diff_filespec_is_binary(two))
1417                goto free_and_return;
1418        else {
1419                /* Crazy xdl interfaces.. */
1420                xpparam_t xpp;
1421                xdemitconf_t xecfg;
1422                xdemitcb_t ecb;
1423
1424                memset(&xecfg, 0, sizeof(xecfg));
1425                xpp.flags = XDF_NEED_MINIMAL;
1426                ecb.outf = xdiff_outf;
1427                ecb.priv = &data;
1428                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1429        }
1430 free_and_return:
1431        diff_free_filespec_data(one);
1432        diff_free_filespec_data(two);
1433}
1434
1435struct diff_filespec *alloc_filespec(const char *path)
1436{
1437        int namelen = strlen(path);
1438        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1439
1440        memset(spec, 0, sizeof(*spec));
1441        spec->path = (char *)(spec + 1);
1442        memcpy(spec->path, path, namelen+1);
1443        spec->count = 1;
1444        return spec;
1445}
1446
1447void free_filespec(struct diff_filespec *spec)
1448{
1449        if (!--spec->count) {
1450                diff_free_filespec_data(spec);
1451                free(spec);
1452        }
1453}
1454
1455void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1456                   unsigned short mode)
1457{
1458        if (mode) {
1459                spec->mode = canon_mode(mode);
1460                hashcpy(spec->sha1, sha1);
1461                spec->sha1_valid = !is_null_sha1(sha1);
1462        }
1463}
1464
1465/*
1466 * Given a name and sha1 pair, if the index tells us the file in
1467 * the work tree has that object contents, return true, so that
1468 * prepare_temp_file() does not have to inflate and extract.
1469 */
1470static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1471{
1472        struct cache_entry *ce;
1473        struct stat st;
1474        int pos, len;
1475
1476        /* We do not read the cache ourselves here, because the
1477         * benchmark with my previous version that always reads cache
1478         * shows that it makes things worse for diff-tree comparing
1479         * two linux-2.6 kernel trees in an already checked out work
1480         * tree.  This is because most diff-tree comparisons deal with
1481         * only a small number of files, while reading the cache is
1482         * expensive for a large project, and its cost outweighs the
1483         * savings we get by not inflating the object to a temporary
1484         * file.  Practically, this code only helps when we are used
1485         * by diff-cache --cached, which does read the cache before
1486         * calling us.
1487         */
1488        if (!active_cache)
1489                return 0;
1490
1491        /* We want to avoid the working directory if our caller
1492         * doesn't need the data in a normal file, this system
1493         * is rather slow with its stat/open/mmap/close syscalls,
1494         * and the object is contained in a pack file.  The pack
1495         * is probably already open and will be faster to obtain
1496         * the data through than the working directory.  Loose
1497         * objects however would tend to be slower as they need
1498         * to be individually opened and inflated.
1499         */
1500        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1501                return 0;
1502
1503        len = strlen(name);
1504        pos = cache_name_pos(name, len);
1505        if (pos < 0)
1506                return 0;
1507        ce = active_cache[pos];
1508        if ((lstat(name, &st) < 0) ||
1509            !S_ISREG(st.st_mode) || /* careful! */
1510            ce_match_stat(ce, &st, 0) ||
1511            hashcmp(sha1, ce->sha1))
1512                return 0;
1513        /* we return 1 only when we can stat, it is a regular file,
1514         * stat information matches, and sha1 recorded in the cache
1515         * matches.  I.e. we know the file in the work tree really is
1516         * the same as the <name, sha1> pair.
1517         */
1518        return 1;
1519}
1520
1521static int populate_from_stdin(struct diff_filespec *s)
1522{
1523        struct strbuf buf;
1524        size_t size = 0;
1525
1526        strbuf_init(&buf, 0);
1527        if (strbuf_read(&buf, 0, 0) < 0)
1528                return error("error while reading from stdin %s",
1529                                     strerror(errno));
1530
1531        s->should_munmap = 0;
1532        s->data = strbuf_detach(&buf, &size);
1533        s->size = size;
1534        s->should_free = 1;
1535        return 0;
1536}
1537
1538static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1539{
1540        int len;
1541        char *data = xmalloc(100);
1542        len = snprintf(data, 100,
1543                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1544        s->data = data;
1545        s->size = len;
1546        s->should_free = 1;
1547        if (size_only) {
1548                s->data = NULL;
1549                free(data);
1550        }
1551        return 0;
1552}
1553
1554/*
1555 * While doing rename detection and pickaxe operation, we may need to
1556 * grab the data for the blob (or file) for our own in-core comparison.
1557 * diff_filespec has data and size fields for this purpose.
1558 */
1559int diff_populate_filespec(struct diff_filespec *s, int size_only)
1560{
1561        int err = 0;
1562        if (!DIFF_FILE_VALID(s))
1563                die("internal error: asking to populate invalid file.");
1564        if (S_ISDIR(s->mode))
1565                return -1;
1566
1567        if (s->data)
1568                return 0;
1569
1570        if (size_only && 0 < s->size)
1571                return 0;
1572
1573        if (S_ISGITLINK(s->mode))
1574                return diff_populate_gitlink(s, size_only);
1575
1576        if (!s->sha1_valid ||
1577            reuse_worktree_file(s->path, s->sha1, 0)) {
1578                struct strbuf buf;
1579                struct stat st;
1580                int fd;
1581
1582                if (!strcmp(s->path, "-"))
1583                        return populate_from_stdin(s);
1584
1585                if (lstat(s->path, &st) < 0) {
1586                        if (errno == ENOENT) {
1587                        err_empty:
1588                                err = -1;
1589                        empty:
1590                                s->data = (char *)"";
1591                                s->size = 0;
1592                                return err;
1593                        }
1594                }
1595                s->size = xsize_t(st.st_size);
1596                if (!s->size)
1597                        goto empty;
1598                if (size_only)
1599                        return 0;
1600                if (S_ISLNK(st.st_mode)) {
1601                        int ret;
1602                        s->data = xmalloc(s->size);
1603                        s->should_free = 1;
1604                        ret = readlink(s->path, s->data, s->size);
1605                        if (ret < 0) {
1606                                free(s->data);
1607                                goto err_empty;
1608                        }
1609                        return 0;
1610                }
1611                fd = open(s->path, O_RDONLY);
1612                if (fd < 0)
1613                        goto err_empty;
1614                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1615                close(fd);
1616                s->should_munmap = 1;
1617
1618                /*
1619                 * Convert from working tree format to canonical git format
1620                 */
1621                strbuf_init(&buf, 0);
1622                if (convert_to_git(s->path, s->data, s->size, &buf)) {
1623                        size_t size = 0;
1624                        munmap(s->data, s->size);
1625                        s->should_munmap = 0;
1626                        s->data = strbuf_detach(&buf, &size);
1627                        s->size = size;
1628                        s->should_free = 1;
1629                }
1630        }
1631        else {
1632                enum object_type type;
1633                if (size_only)
1634                        type = sha1_object_info(s->sha1, &s->size);
1635                else {
1636                        s->data = read_sha1_file(s->sha1, &type, &s->size);
1637                        s->should_free = 1;
1638                }
1639        }
1640        return 0;
1641}
1642
1643void diff_free_filespec_blob(struct diff_filespec *s)
1644{
1645        if (s->should_free)
1646                free(s->data);
1647        else if (s->should_munmap)
1648                munmap(s->data, s->size);
1649
1650        if (s->should_free || s->should_munmap) {
1651                s->should_free = s->should_munmap = 0;
1652                s->data = NULL;
1653        }
1654}
1655
1656void diff_free_filespec_data(struct diff_filespec *s)
1657{
1658        diff_free_filespec_blob(s);
1659        free(s->cnt_data);
1660        s->cnt_data = NULL;
1661}
1662
1663static void prep_temp_blob(struct diff_tempfile *temp,
1664                           void *blob,
1665                           unsigned long size,
1666                           const unsigned char *sha1,
1667                           int mode)
1668{
1669        int fd;
1670
1671        fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1672        if (fd < 0)
1673                die("unable to create temp-file: %s", strerror(errno));
1674        if (write_in_full(fd, blob, size) != size)
1675                die("unable to write temp-file");
1676        close(fd);
1677        temp->name = temp->tmp_path;
1678        strcpy(temp->hex, sha1_to_hex(sha1));
1679        temp->hex[40] = 0;
1680        sprintf(temp->mode, "%06o", mode);
1681}
1682
1683static void prepare_temp_file(const char *name,
1684                              struct diff_tempfile *temp,
1685                              struct diff_filespec *one)
1686{
1687        if (!DIFF_FILE_VALID(one)) {
1688        not_a_valid_file:
1689                /* A '-' entry produces this for file-2, and
1690                 * a '+' entry produces this for file-1.
1691                 */
1692                temp->name = "/dev/null";
1693                strcpy(temp->hex, ".");
1694                strcpy(temp->mode, ".");
1695                return;
1696        }
1697
1698        if (!one->sha1_valid ||
1699            reuse_worktree_file(name, one->sha1, 1)) {
1700                struct stat st;
1701                if (lstat(name, &st) < 0) {
1702                        if (errno == ENOENT)
1703                                goto not_a_valid_file;
1704                        die("stat(%s): %s", name, strerror(errno));
1705                }
1706                if (S_ISLNK(st.st_mode)) {
1707                        int ret;
1708                        char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1709                        size_t sz = xsize_t(st.st_size);
1710                        if (sizeof(buf) <= st.st_size)
1711                                die("symlink too long: %s", name);
1712                        ret = readlink(name, buf, sz);
1713                        if (ret < 0)
1714                                die("readlink(%s)", name);
1715                        prep_temp_blob(temp, buf, sz,
1716                                       (one->sha1_valid ?
1717                                        one->sha1 : null_sha1),
1718                                       (one->sha1_valid ?
1719                                        one->mode : S_IFLNK));
1720                }
1721                else {
1722                        /* we can borrow from the file in the work tree */
1723                        temp->name = name;
1724                        if (!one->sha1_valid)
1725                                strcpy(temp->hex, sha1_to_hex(null_sha1));
1726                        else
1727                                strcpy(temp->hex, sha1_to_hex(one->sha1));
1728                        /* Even though we may sometimes borrow the
1729                         * contents from the work tree, we always want
1730                         * one->mode.  mode is trustworthy even when
1731                         * !(one->sha1_valid), as long as
1732                         * DIFF_FILE_VALID(one).
1733                         */
1734                        sprintf(temp->mode, "%06o", one->mode);
1735                }
1736                return;
1737        }
1738        else {
1739                if (diff_populate_filespec(one, 0))
1740                        die("cannot read data blob for %s", one->path);
1741                prep_temp_blob(temp, one->data, one->size,
1742                               one->sha1, one->mode);
1743        }
1744}
1745
1746static void remove_tempfile(void)
1747{
1748        int i;
1749
1750        for (i = 0; i < 2; i++)
1751                if (diff_temp[i].name == diff_temp[i].tmp_path) {
1752                        unlink(diff_temp[i].name);
1753                        diff_temp[i].name = NULL;
1754                }
1755}
1756
1757static void remove_tempfile_on_signal(int signo)
1758{
1759        remove_tempfile();
1760        signal(SIGINT, SIG_DFL);
1761        raise(signo);
1762}
1763
1764static int spawn_prog(const char *pgm, const char **arg)
1765{
1766        pid_t pid;
1767        int status;
1768
1769        fflush(NULL);
1770        pid = fork();
1771        if (pid < 0)
1772                die("unable to fork");
1773        if (!pid) {
1774                execvp(pgm, (char *const*) arg);
1775                exit(255);
1776        }
1777
1778        while (waitpid(pid, &status, 0) < 0) {
1779                if (errno == EINTR)
1780                        continue;
1781                return -1;
1782        }
1783
1784        /* Earlier we did not check the exit status because
1785         * diff exits non-zero if files are different, and
1786         * we are not interested in knowing that.  It was a
1787         * mistake which made it harder to quit a diff-*
1788         * session that uses the git-apply-patch-script as
1789         * the GIT_EXTERNAL_DIFF.  A custom GIT_EXTERNAL_DIFF
1790         * should also exit non-zero only when it wants to
1791         * abort the entire diff-* session.
1792         */
1793        if (WIFEXITED(status) && !WEXITSTATUS(status))
1794                return 0;
1795        return -1;
1796}
1797
1798/* An external diff command takes:
1799 *
1800 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1801 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
1802 *
1803 */
1804static void run_external_diff(const char *pgm,
1805                              const char *name,
1806                              const char *other,
1807                              struct diff_filespec *one,
1808                              struct diff_filespec *two,
1809                              const char *xfrm_msg,
1810                              int complete_rewrite)
1811{
1812        const char *spawn_arg[10];
1813        struct diff_tempfile *temp = diff_temp;
1814        int retval;
1815        static int atexit_asked = 0;
1816        const char *othername;
1817        const char **arg = &spawn_arg[0];
1818
1819        othername = (other? other : name);
1820        if (one && two) {
1821                prepare_temp_file(name, &temp[0], one);
1822                prepare_temp_file(othername, &temp[1], two);
1823                if (! atexit_asked &&
1824                    (temp[0].name == temp[0].tmp_path ||
1825                     temp[1].name == temp[1].tmp_path)) {
1826                        atexit_asked = 1;
1827                        atexit(remove_tempfile);
1828                }
1829                signal(SIGINT, remove_tempfile_on_signal);
1830        }
1831
1832        if (one && two) {
1833                *arg++ = pgm;
1834                *arg++ = name;
1835                *arg++ = temp[0].name;
1836                *arg++ = temp[0].hex;
1837                *arg++ = temp[0].mode;
1838                *arg++ = temp[1].name;
1839                *arg++ = temp[1].hex;
1840                *arg++ = temp[1].mode;
1841                if (other) {
1842                        *arg++ = other;
1843                        *arg++ = xfrm_msg;
1844                }
1845        } else {
1846                *arg++ = pgm;
1847                *arg++ = name;
1848        }
1849        *arg = NULL;
1850        retval = spawn_prog(pgm, spawn_arg);
1851        remove_tempfile();
1852        if (retval) {
1853                fprintf(stderr, "external diff died, stopping at %s.\n", name);
1854                exit(1);
1855        }
1856}
1857
1858static const char *external_diff_attr(const char *name)
1859{
1860        struct git_attr_check attr_diff_check;
1861
1862        setup_diff_attr_check(&attr_diff_check);
1863        if (!git_checkattr(name, 1, &attr_diff_check)) {
1864                const char *value = attr_diff_check.value;
1865                if (!ATTR_TRUE(value) &&
1866                    !ATTR_FALSE(value) &&
1867                    !ATTR_UNSET(value)) {
1868                        struct ll_diff_driver *drv;
1869
1870                        read_config_if_needed();
1871                        for (drv = user_diff; drv; drv = drv->next)
1872                                if (!strcmp(drv->name, value))
1873                                        return drv->cmd;
1874                }
1875        }
1876        return NULL;
1877}
1878
1879static void run_diff_cmd(const char *pgm,
1880                         const char *name,
1881                         const char *other,
1882                         struct diff_filespec *one,
1883                         struct diff_filespec *two,
1884                         const char *xfrm_msg,
1885                         struct diff_options *o,
1886                         int complete_rewrite)
1887{
1888        if (!o->allow_external)
1889                pgm = NULL;
1890        else {
1891                const char *cmd = external_diff_attr(name);
1892                if (cmd)
1893                        pgm = cmd;
1894        }
1895
1896        if (pgm) {
1897                run_external_diff(pgm, name, other, one, two, xfrm_msg,
1898                                  complete_rewrite);
1899                return;
1900        }
1901        if (one && two)
1902                builtin_diff(name, other ? other : name,
1903                             one, two, xfrm_msg, o, complete_rewrite);
1904        else
1905                printf("* Unmerged path %s\n", name);
1906}
1907
1908static void diff_fill_sha1_info(struct diff_filespec *one)
1909{
1910        if (DIFF_FILE_VALID(one)) {
1911                if (!one->sha1_valid) {
1912                        struct stat st;
1913                        if (!strcmp(one->path, "-")) {
1914                                hashcpy(one->sha1, null_sha1);
1915                                return;
1916                        }
1917                        if (lstat(one->path, &st) < 0)
1918                                die("stat %s", one->path);
1919                        if (index_path(one->sha1, one->path, &st, 0))
1920                                die("cannot hash %s\n", one->path);
1921                }
1922        }
1923        else
1924                hashclr(one->sha1);
1925}
1926
1927static int similarity_index(struct diff_filepair *p)
1928{
1929        return p->score * 100 / MAX_SCORE;
1930}
1931
1932static void run_diff(struct diff_filepair *p, struct diff_options *o)
1933{
1934        const char *pgm = external_diff();
1935        struct strbuf msg;
1936        char *xfrm_msg;
1937        struct diff_filespec *one = p->one;
1938        struct diff_filespec *two = p->two;
1939        const char *name;
1940        const char *other;
1941        int complete_rewrite = 0;
1942
1943
1944        if (DIFF_PAIR_UNMERGED(p)) {
1945                run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1946                return;
1947        }
1948
1949        name  = p->one->path;
1950        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1951        diff_fill_sha1_info(one);
1952        diff_fill_sha1_info(two);
1953
1954        strbuf_init(&msg, PATH_MAX * 2 + 300);
1955        switch (p->status) {
1956        case DIFF_STATUS_COPIED:
1957                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1958                strbuf_addstr(&msg, "\ncopy from ");
1959                quote_c_style(name, &msg, NULL, 0);
1960                strbuf_addstr(&msg, "\ncopy to ");
1961                quote_c_style(other, &msg, NULL, 0);
1962                strbuf_addch(&msg, '\n');
1963                break;
1964        case DIFF_STATUS_RENAMED:
1965                strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
1966                strbuf_addstr(&msg, "\nrename from ");
1967                quote_c_style(name, &msg, NULL, 0);
1968                strbuf_addstr(&msg, "\nrename to ");
1969                quote_c_style(other, &msg, NULL, 0);
1970                strbuf_addch(&msg, '\n');
1971                break;
1972        case DIFF_STATUS_MODIFIED:
1973                if (p->score) {
1974                        strbuf_addf(&msg, "dissimilarity index %d%%\n",
1975                                        similarity_index(p));
1976                        complete_rewrite = 1;
1977                        break;
1978                }
1979                /* fallthru */
1980        default:
1981                /* nothing */
1982                ;
1983        }
1984
1985        if (hashcmp(one->sha1, two->sha1)) {
1986                int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
1987
1988                if (o->binary) {
1989                        mmfile_t mf;
1990                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
1991                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
1992                                abbrev = 40;
1993                }
1994                strbuf_addf(&msg, "index %.*s..%.*s",
1995                                abbrev, sha1_to_hex(one->sha1),
1996                                abbrev, sha1_to_hex(two->sha1));
1997                if (one->mode == two->mode)
1998                        strbuf_addf(&msg, " %06o", one->mode);
1999                strbuf_addch(&msg, '\n');
2000        }
2001
2002        if (msg.len)
2003                strbuf_setlen(&msg, msg.len - 1);
2004        xfrm_msg = msg.len ? msg.buf : NULL;
2005
2006        if (!pgm &&
2007            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2008            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2009                /* a filepair that changes between file and symlink
2010                 * needs to be split into deletion and creation.
2011                 */
2012                struct diff_filespec *null = alloc_filespec(two->path);
2013                run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
2014                free(null);
2015                null = alloc_filespec(one->path);
2016                run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
2017                free(null);
2018        }
2019        else
2020                run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
2021                             complete_rewrite);
2022
2023        strbuf_release(&msg);
2024}
2025
2026static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2027                         struct diffstat_t *diffstat)
2028{
2029        const char *name;
2030        const char *other;
2031        int complete_rewrite = 0;
2032
2033        if (DIFF_PAIR_UNMERGED(p)) {
2034                /* unmerged */
2035                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2036                return;
2037        }
2038
2039        name = p->one->path;
2040        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2041
2042        diff_fill_sha1_info(p->one);
2043        diff_fill_sha1_info(p->two);
2044
2045        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2046                complete_rewrite = 1;
2047        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2048}
2049
2050static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2051{
2052        const char *name;
2053        const char *other;
2054
2055        if (DIFF_PAIR_UNMERGED(p)) {
2056                /* unmerged */
2057                return;
2058        }
2059
2060        name = p->one->path;
2061        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2062
2063        diff_fill_sha1_info(p->one);
2064        diff_fill_sha1_info(p->two);
2065
2066        builtin_checkdiff(name, other, p->one, p->two, o);
2067}
2068
2069void diff_setup(struct diff_options *options)
2070{
2071        memset(options, 0, sizeof(*options));
2072        options->line_termination = '\n';
2073        options->break_opt = -1;
2074        options->rename_limit = -1;
2075        options->context = 3;
2076        options->msg_sep = "";
2077
2078        options->change = diff_change;
2079        options->add_remove = diff_addremove;
2080        options->color_diff = diff_use_color_default;
2081        options->detect_rename = diff_detect_rename_default;
2082}
2083
2084int diff_setup_done(struct diff_options *options)
2085{
2086        int count = 0;
2087
2088        if (options->output_format & DIFF_FORMAT_NAME)
2089                count++;
2090        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2091                count++;
2092        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2093                count++;
2094        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2095                count++;
2096        if (count > 1)
2097                die("--name-only, --name-status, --check and -s are mutually exclusive");
2098
2099        if (options->find_copies_harder)
2100                options->detect_rename = DIFF_DETECT_COPY;
2101
2102        if (options->output_format & (DIFF_FORMAT_NAME |
2103                                      DIFF_FORMAT_NAME_STATUS |
2104                                      DIFF_FORMAT_CHECKDIFF |
2105                                      DIFF_FORMAT_NO_OUTPUT))
2106                options->output_format &= ~(DIFF_FORMAT_RAW |
2107                                            DIFF_FORMAT_NUMSTAT |
2108                                            DIFF_FORMAT_DIFFSTAT |
2109                                            DIFF_FORMAT_SHORTSTAT |
2110                                            DIFF_FORMAT_SUMMARY |
2111                                            DIFF_FORMAT_PATCH);
2112
2113        /*
2114         * These cases always need recursive; we do not drop caller-supplied
2115         * recursive bits for other formats here.
2116         */
2117        if (options->output_format & (DIFF_FORMAT_PATCH |
2118                                      DIFF_FORMAT_NUMSTAT |
2119                                      DIFF_FORMAT_DIFFSTAT |
2120                                      DIFF_FORMAT_SHORTSTAT |
2121                                      DIFF_FORMAT_SUMMARY |
2122                                      DIFF_FORMAT_CHECKDIFF))
2123                options->recursive = 1;
2124        /*
2125         * Also pickaxe would not work very well if you do not say recursive
2126         */
2127        if (options->pickaxe)
2128                options->recursive = 1;
2129
2130        if (options->detect_rename && options->rename_limit < 0)
2131                options->rename_limit = diff_rename_limit_default;
2132        if (options->setup & DIFF_SETUP_USE_CACHE) {
2133                if (!active_cache)
2134                        /* read-cache does not die even when it fails
2135                         * so it is safe for us to do this here.  Also
2136                         * it does not smudge active_cache or active_nr
2137                         * when it fails, so we do not have to worry about
2138                         * cleaning it up ourselves either.
2139                         */
2140                        read_cache();
2141        }
2142        if (options->abbrev <= 0 || 40 < options->abbrev)
2143                options->abbrev = 40; /* full */
2144
2145        /*
2146         * It does not make sense to show the first hit we happened
2147         * to have found.  It does not make sense not to return with
2148         * exit code in such a case either.
2149         */
2150        if (options->quiet) {
2151                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2152                options->exit_with_status = 1;
2153        }
2154
2155        /*
2156         * If we postprocess in diffcore, we cannot simply return
2157         * upon the first hit.  We need to run diff as usual.
2158         */
2159        if (options->pickaxe || options->filter)
2160                options->quiet = 0;
2161
2162        return 0;
2163}
2164
2165static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2166{
2167        char c, *eq;
2168        int len;
2169
2170        if (*arg != '-')
2171                return 0;
2172        c = *++arg;
2173        if (!c)
2174                return 0;
2175        if (c == arg_short) {
2176                c = *++arg;
2177                if (!c)
2178                        return 1;
2179                if (val && isdigit(c)) {
2180                        char *end;
2181                        int n = strtoul(arg, &end, 10);
2182                        if (*end)
2183                                return 0;
2184                        *val = n;
2185                        return 1;
2186                }
2187                return 0;
2188        }
2189        if (c != '-')
2190                return 0;
2191        arg++;
2192        eq = strchr(arg, '=');
2193        if (eq)
2194                len = eq - arg;
2195        else
2196                len = strlen(arg);
2197        if (!len || strncmp(arg, arg_long, len))
2198                return 0;
2199        if (eq) {
2200                int n;
2201                char *end;
2202                if (!isdigit(*++eq))
2203                        return 0;
2204                n = strtoul(eq, &end, 10);
2205                if (*end)
2206                        return 0;
2207                *val = n;
2208        }
2209        return 1;
2210}
2211
2212static int diff_scoreopt_parse(const char *opt);
2213
2214int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2215{
2216        const char *arg = av[0];
2217        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2218                options->output_format |= DIFF_FORMAT_PATCH;
2219        else if (opt_arg(arg, 'U', "unified", &options->context))
2220                options->output_format |= DIFF_FORMAT_PATCH;
2221        else if (!strcmp(arg, "--raw"))
2222                options->output_format |= DIFF_FORMAT_RAW;
2223        else if (!strcmp(arg, "--patch-with-raw")) {
2224                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2225        }
2226        else if (!strcmp(arg, "--numstat")) {
2227                options->output_format |= DIFF_FORMAT_NUMSTAT;
2228        }
2229        else if (!strcmp(arg, "--shortstat")) {
2230                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2231        }
2232        else if (!prefixcmp(arg, "--stat")) {
2233                char *end;
2234                int width = options->stat_width;
2235                int name_width = options->stat_name_width;
2236                arg += 6;
2237                end = (char *)arg;
2238
2239                switch (*arg) {
2240                case '-':
2241                        if (!prefixcmp(arg, "-width="))
2242                                width = strtoul(arg + 7, &end, 10);
2243                        else if (!prefixcmp(arg, "-name-width="))
2244                                name_width = strtoul(arg + 12, &end, 10);
2245                        break;
2246                case '=':
2247                        width = strtoul(arg+1, &end, 10);
2248                        if (*end == ',')
2249                                name_width = strtoul(end+1, &end, 10);
2250                }
2251
2252                /* Important! This checks all the error cases! */
2253                if (*end)
2254                        return 0;
2255                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2256                options->stat_name_width = name_width;
2257                options->stat_width = width;
2258        }
2259        else if (!strcmp(arg, "--check"))
2260                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2261        else if (!strcmp(arg, "--summary"))
2262                options->output_format |= DIFF_FORMAT_SUMMARY;
2263        else if (!strcmp(arg, "--patch-with-stat")) {
2264                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2265        }
2266        else if (!strcmp(arg, "-z"))
2267                options->line_termination = 0;
2268        else if (!prefixcmp(arg, "-l"))
2269                options->rename_limit = strtoul(arg+2, NULL, 10);
2270        else if (!strcmp(arg, "--full-index"))
2271                options->full_index = 1;
2272        else if (!strcmp(arg, "--binary")) {
2273                options->output_format |= DIFF_FORMAT_PATCH;
2274                options->binary = 1;
2275        }
2276        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text")) {
2277                options->text = 1;
2278        }
2279        else if (!strcmp(arg, "--name-only"))
2280                options->output_format |= DIFF_FORMAT_NAME;
2281        else if (!strcmp(arg, "--name-status"))
2282                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2283        else if (!strcmp(arg, "-R"))
2284                options->reverse_diff = 1;
2285        else if (!prefixcmp(arg, "-S"))
2286                options->pickaxe = arg + 2;
2287        else if (!strcmp(arg, "-s")) {
2288                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2289        }
2290        else if (!prefixcmp(arg, "-O"))
2291                options->orderfile = arg + 2;
2292        else if (!prefixcmp(arg, "--diff-filter="))
2293                options->filter = arg + 14;
2294        else if (!strcmp(arg, "--pickaxe-all"))
2295                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2296        else if (!strcmp(arg, "--pickaxe-regex"))
2297                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2298        else if (!prefixcmp(arg, "-B")) {
2299                if ((options->break_opt =
2300                     diff_scoreopt_parse(arg)) == -1)
2301                        return -1;
2302        }
2303        else if (!prefixcmp(arg, "-M")) {
2304                if ((options->rename_score =
2305                     diff_scoreopt_parse(arg)) == -1)
2306                        return -1;
2307                options->detect_rename = DIFF_DETECT_RENAME;
2308        }
2309        else if (!prefixcmp(arg, "-C")) {
2310                if (options->detect_rename == DIFF_DETECT_COPY)
2311                        options->find_copies_harder = 1;
2312                if ((options->rename_score =
2313                     diff_scoreopt_parse(arg)) == -1)
2314                        return -1;
2315                options->detect_rename = DIFF_DETECT_COPY;
2316        }
2317        else if (!strcmp(arg, "--find-copies-harder"))
2318                options->find_copies_harder = 1;
2319        else if (!strcmp(arg, "--follow"))
2320                options->follow_renames = 1;
2321        else if (!strcmp(arg, "--abbrev"))
2322                options->abbrev = DEFAULT_ABBREV;
2323        else if (!prefixcmp(arg, "--abbrev=")) {
2324                options->abbrev = strtoul(arg + 9, NULL, 10);
2325                if (options->abbrev < MINIMUM_ABBREV)
2326                        options->abbrev = MINIMUM_ABBREV;
2327                else if (40 < options->abbrev)
2328                        options->abbrev = 40;
2329        }
2330        else if (!strcmp(arg, "--color"))
2331                options->color_diff = 1;
2332        else if (!strcmp(arg, "--no-color"))
2333                options->color_diff = 0;
2334        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2335                options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2336        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2337                options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2338        else if (!strcmp(arg, "--ignore-space-at-eol"))
2339                options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2340        else if (!strcmp(arg, "--color-words"))
2341                options->color_diff = options->color_diff_words = 1;
2342        else if (!strcmp(arg, "--no-renames"))
2343                options->detect_rename = 0;
2344        else if (!strcmp(arg, "--exit-code"))
2345                options->exit_with_status = 1;
2346        else if (!strcmp(arg, "--quiet"))
2347                options->quiet = 1;
2348        else if (!strcmp(arg, "--ext-diff"))
2349                options->allow_external = 1;
2350        else if (!strcmp(arg, "--no-ext-diff"))
2351                options->allow_external = 0;
2352        else
2353                return 0;
2354        return 1;
2355}
2356
2357static int parse_num(const char **cp_p)
2358{
2359        unsigned long num, scale;
2360        int ch, dot;
2361        const char *cp = *cp_p;
2362
2363        num = 0;
2364        scale = 1;
2365        dot = 0;
2366        for(;;) {
2367                ch = *cp;
2368                if ( !dot && ch == '.' ) {
2369                        scale = 1;
2370                        dot = 1;
2371                } else if ( ch == '%' ) {
2372                        scale = dot ? scale*100 : 100;
2373                        cp++;   /* % is always at the end */
2374                        break;
2375                } else if ( ch >= '0' && ch <= '9' ) {
2376                        if ( scale < 100000 ) {
2377                                scale *= 10;
2378                                num = (num*10) + (ch-'0');
2379                        }
2380                } else {
2381                        break;
2382                }
2383                cp++;
2384        }
2385        *cp_p = cp;
2386
2387        /* user says num divided by scale and we say internally that
2388         * is MAX_SCORE * num / scale.
2389         */
2390        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2391}
2392
2393static int diff_scoreopt_parse(const char *opt)
2394{
2395        int opt1, opt2, cmd;
2396
2397        if (*opt++ != '-')
2398                return -1;
2399        cmd = *opt++;
2400        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2401                return -1; /* that is not a -M, -C nor -B option */
2402
2403        opt1 = parse_num(&opt);
2404        if (cmd != 'B')
2405                opt2 = 0;
2406        else {
2407                if (*opt == 0)
2408                        opt2 = 0;
2409                else if (*opt != '/')
2410                        return -1; /* we expect -B80/99 or -B80 */
2411                else {
2412                        opt++;
2413                        opt2 = parse_num(&opt);
2414                }
2415        }
2416        if (*opt != 0)
2417                return -1;
2418        return opt1 | (opt2 << 16);
2419}
2420
2421struct diff_queue_struct diff_queued_diff;
2422
2423void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2424{
2425        if (queue->alloc <= queue->nr) {
2426                queue->alloc = alloc_nr(queue->alloc);
2427                queue->queue = xrealloc(queue->queue,
2428                                        sizeof(dp) * queue->alloc);
2429        }
2430        queue->queue[queue->nr++] = dp;
2431}
2432
2433struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2434                                 struct diff_filespec *one,
2435                                 struct diff_filespec *two)
2436{
2437        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2438        dp->one = one;
2439        dp->two = two;
2440        if (queue)
2441                diff_q(queue, dp);
2442        return dp;
2443}
2444
2445void diff_free_filepair(struct diff_filepair *p)
2446{
2447        free_filespec(p->one);
2448        free_filespec(p->two);
2449        free(p);
2450}
2451
2452/* This is different from find_unique_abbrev() in that
2453 * it stuffs the result with dots for alignment.
2454 */
2455const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2456{
2457        int abblen;
2458        const char *abbrev;
2459        if (len == 40)
2460                return sha1_to_hex(sha1);
2461
2462        abbrev = find_unique_abbrev(sha1, len);
2463        if (!abbrev)
2464                return sha1_to_hex(sha1);
2465        abblen = strlen(abbrev);
2466        if (abblen < 37) {
2467                static char hex[41];
2468                if (len < abblen && abblen <= len + 2)
2469                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2470                else
2471                        sprintf(hex, "%s...", abbrev);
2472                return hex;
2473        }
2474        return sha1_to_hex(sha1);
2475}
2476
2477static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2478{
2479        int line_termination = opt->line_termination;
2480        int inter_name_termination = line_termination ? '\t' : '\0';
2481
2482        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2483                printf(":%06o %06o %s ", p->one->mode, p->two->mode,
2484                       diff_unique_abbrev(p->one->sha1, opt->abbrev));
2485                printf("%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2486        }
2487        if (p->score) {
2488                printf("%c%03d%c", p->status, similarity_index(p),
2489                           inter_name_termination);
2490        } else {
2491                printf("%c%c", p->status, inter_name_termination);
2492        }
2493
2494        if (p->status == DIFF_STATUS_COPIED || p->status == DIFF_STATUS_RENAMED) {
2495                write_name_quoted(p->one->path, stdout, inter_name_termination);
2496                write_name_quoted(p->two->path, stdout, line_termination);
2497        } else {
2498                const char *path = p->one->mode ? p->one->path : p->two->path;
2499                write_name_quoted(path, stdout, line_termination);
2500        }
2501}
2502
2503int diff_unmodified_pair(struct diff_filepair *p)
2504{
2505        /* This function is written stricter than necessary to support
2506         * the currently implemented transformers, but the idea is to
2507         * let transformers to produce diff_filepairs any way they want,
2508         * and filter and clean them up here before producing the output.
2509         */
2510        struct diff_filespec *one = p->one, *two = p->two;
2511
2512        if (DIFF_PAIR_UNMERGED(p))
2513                return 0; /* unmerged is interesting */
2514
2515        /* deletion, addition, mode or type change
2516         * and rename are all interesting.
2517         */
2518        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2519            DIFF_PAIR_MODE_CHANGED(p) ||
2520            strcmp(one->path, two->path))
2521                return 0;
2522
2523        /* both are valid and point at the same path.  that is, we are
2524         * dealing with a change.
2525         */
2526        if (one->sha1_valid && two->sha1_valid &&
2527            !hashcmp(one->sha1, two->sha1))
2528                return 1; /* no change */
2529        if (!one->sha1_valid && !two->sha1_valid)
2530                return 1; /* both look at the same file on the filesystem. */
2531        return 0;
2532}
2533
2534static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2535{
2536        if (diff_unmodified_pair(p))
2537                return;
2538
2539        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2540            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2541                return; /* no tree diffs in patch format */
2542
2543        run_diff(p, o);
2544}
2545
2546static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2547                            struct diffstat_t *diffstat)
2548{
2549        if (diff_unmodified_pair(p))
2550                return;
2551
2552        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2553            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2554                return; /* no tree diffs in patch format */
2555
2556        run_diffstat(p, o, diffstat);
2557}
2558
2559static void diff_flush_checkdiff(struct diff_filepair *p,
2560                struct diff_options *o)
2561{
2562        if (diff_unmodified_pair(p))
2563                return;
2564
2565        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2566            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2567                return; /* no tree diffs in patch format */
2568
2569        run_checkdiff(p, o);
2570}
2571
2572int diff_queue_is_empty(void)
2573{
2574        struct diff_queue_struct *q = &diff_queued_diff;
2575        int i;
2576        for (i = 0; i < q->nr; i++)
2577                if (!diff_unmodified_pair(q->queue[i]))
2578                        return 0;
2579        return 1;
2580}
2581
2582#if DIFF_DEBUG
2583void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2584{
2585        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2586                x, one ? one : "",
2587                s->path,
2588                DIFF_FILE_VALID(s) ? "valid" : "invalid",
2589                s->mode,
2590                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2591        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2592                x, one ? one : "",
2593                s->size, s->xfrm_flags);
2594}
2595
2596void diff_debug_filepair(const struct diff_filepair *p, int i)
2597{
2598        diff_debug_filespec(p->one, i, "one");
2599        diff_debug_filespec(p->two, i, "two");
2600        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2601                p->score, p->status ? p->status : '?',
2602                p->one->rename_used, p->broken_pair);
2603}
2604
2605void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2606{
2607        int i;
2608        if (msg)
2609                fprintf(stderr, "%s\n", msg);
2610        fprintf(stderr, "q->nr = %d\n", q->nr);
2611        for (i = 0; i < q->nr; i++) {
2612                struct diff_filepair *p = q->queue[i];
2613                diff_debug_filepair(p, i);
2614        }
2615}
2616#endif
2617
2618static void diff_resolve_rename_copy(void)
2619{
2620        int i;
2621        struct diff_filepair *p;
2622        struct diff_queue_struct *q = &diff_queued_diff;
2623
2624        diff_debug_queue("resolve-rename-copy", q);
2625
2626        for (i = 0; i < q->nr; i++) {
2627                p = q->queue[i];
2628                p->status = 0; /* undecided */
2629                if (DIFF_PAIR_UNMERGED(p))
2630                        p->status = DIFF_STATUS_UNMERGED;
2631                else if (!DIFF_FILE_VALID(p->one))
2632                        p->status = DIFF_STATUS_ADDED;
2633                else if (!DIFF_FILE_VALID(p->two))
2634                        p->status = DIFF_STATUS_DELETED;
2635                else if (DIFF_PAIR_TYPE_CHANGED(p))
2636                        p->status = DIFF_STATUS_TYPE_CHANGED;
2637
2638                /* from this point on, we are dealing with a pair
2639                 * whose both sides are valid and of the same type, i.e.
2640                 * either in-place edit or rename/copy edit.
2641                 */
2642                else if (DIFF_PAIR_RENAME(p)) {
2643                        /*
2644                         * A rename might have re-connected a broken
2645                         * pair up, causing the pathnames to be the
2646                         * same again. If so, that's not a rename at
2647                         * all, just a modification..
2648                         *
2649                         * Otherwise, see if this source was used for
2650                         * multiple renames, in which case we decrement
2651                         * the count, and call it a copy.
2652                         */
2653                        if (!strcmp(p->one->path, p->two->path))
2654                                p->status = DIFF_STATUS_MODIFIED;
2655                        else if (--p->one->rename_used > 0)
2656                                p->status = DIFF_STATUS_COPIED;
2657                        else
2658                                p->status = DIFF_STATUS_RENAMED;
2659                }
2660                else if (hashcmp(p->one->sha1, p->two->sha1) ||
2661                         p->one->mode != p->two->mode ||
2662                         is_null_sha1(p->one->sha1))
2663                        p->status = DIFF_STATUS_MODIFIED;
2664                else {
2665                        /* This is a "no-change" entry and should not
2666                         * happen anymore, but prepare for broken callers.
2667                         */
2668                        error("feeding unmodified %s to diffcore",
2669                              p->one->path);
2670                        p->status = DIFF_STATUS_UNKNOWN;
2671                }
2672        }
2673        diff_debug_queue("resolve-rename-copy done", q);
2674}
2675
2676static int check_pair_status(struct diff_filepair *p)
2677{
2678        switch (p->status) {
2679        case DIFF_STATUS_UNKNOWN:
2680                return 0;
2681        case 0:
2682                die("internal error in diff-resolve-rename-copy");
2683        default:
2684                return 1;
2685        }
2686}
2687
2688static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2689{
2690        int fmt = opt->output_format;
2691
2692        if (fmt & DIFF_FORMAT_CHECKDIFF)
2693                diff_flush_checkdiff(p, opt);
2694        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2695                diff_flush_raw(p, opt);
2696        else if (fmt & DIFF_FORMAT_NAME)
2697                write_name_quoted(p->two->path, stdout, opt->line_termination);
2698}
2699
2700static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2701{
2702        if (fs->mode)
2703                printf(" %s mode %06o ", newdelete, fs->mode);
2704        else
2705                printf(" %s ", newdelete);
2706        write_name_quoted(fs->path, stdout, '\n');
2707}
2708
2709
2710static void show_mode_change(struct diff_filepair *p, int show_name)
2711{
2712        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2713                printf(" mode change %06o => %06o%c", p->one->mode, p->two->mode,
2714                        show_name ? ' ' : '\n');
2715                if (show_name) {
2716                        write_name_quoted(p->two->path, stdout, '\n');
2717                }
2718        }
2719}
2720
2721static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2722{
2723        char *names = pprint_rename(p->one->path, p->two->path);
2724
2725        printf(" %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2726        free(names);
2727        show_mode_change(p, 0);
2728}
2729
2730static void diff_summary(struct diff_filepair *p)
2731{
2732        switch(p->status) {
2733        case DIFF_STATUS_DELETED:
2734                show_file_mode_name("delete", p->one);
2735                break;
2736        case DIFF_STATUS_ADDED:
2737                show_file_mode_name("create", p->two);
2738                break;
2739        case DIFF_STATUS_COPIED:
2740                show_rename_copy("copy", p);
2741                break;
2742        case DIFF_STATUS_RENAMED:
2743                show_rename_copy("rename", p);
2744                break;
2745        default:
2746                if (p->score) {
2747                        puts(" rewrite ");
2748                        write_name_quoted(p->two->path, stdout, ' ');
2749                        printf("(%d%%)\n", similarity_index(p));
2750                }
2751                show_mode_change(p, !p->score);
2752                break;
2753        }
2754}
2755
2756struct patch_id_t {
2757        struct xdiff_emit_state xm;
2758        SHA_CTX *ctx;
2759        int patchlen;
2760};
2761
2762static int remove_space(char *line, int len)
2763{
2764        int i;
2765        char *dst = line;
2766        unsigned char c;
2767
2768        for (i = 0; i < len; i++)
2769                if (!isspace((c = line[i])))
2770                        *dst++ = c;
2771
2772        return dst - line;
2773}
2774
2775static void patch_id_consume(void *priv, char *line, unsigned long len)
2776{
2777        struct patch_id_t *data = priv;
2778        int new_len;
2779
2780        /* Ignore line numbers when computing the SHA1 of the patch */
2781        if (!prefixcmp(line, "@@ -"))
2782                return;
2783
2784        new_len = remove_space(line, len);
2785
2786        SHA1_Update(data->ctx, line, new_len);
2787        data->patchlen += new_len;
2788}
2789
2790/* returns 0 upon success, and writes result into sha1 */
2791static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2792{
2793        struct diff_queue_struct *q = &diff_queued_diff;
2794        int i;
2795        SHA_CTX ctx;
2796        struct patch_id_t data;
2797        char buffer[PATH_MAX * 4 + 20];
2798
2799        SHA1_Init(&ctx);
2800        memset(&data, 0, sizeof(struct patch_id_t));
2801        data.ctx = &ctx;
2802        data.xm.consume = patch_id_consume;
2803
2804        for (i = 0; i < q->nr; i++) {
2805                xpparam_t xpp;
2806                xdemitconf_t xecfg;
2807                xdemitcb_t ecb;
2808                mmfile_t mf1, mf2;
2809                struct diff_filepair *p = q->queue[i];
2810                int len1, len2;
2811
2812                memset(&xecfg, 0, sizeof(xecfg));
2813                if (p->status == 0)
2814                        return error("internal diff status error");
2815                if (p->status == DIFF_STATUS_UNKNOWN)
2816                        continue;
2817                if (diff_unmodified_pair(p))
2818                        continue;
2819                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2820                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2821                        continue;
2822                if (DIFF_PAIR_UNMERGED(p))
2823                        continue;
2824
2825                diff_fill_sha1_info(p->one);
2826                diff_fill_sha1_info(p->two);
2827                if (fill_mmfile(&mf1, p->one) < 0 ||
2828                                fill_mmfile(&mf2, p->two) < 0)
2829                        return error("unable to read files to diff");
2830
2831                len1 = remove_space(p->one->path, strlen(p->one->path));
2832                len2 = remove_space(p->two->path, strlen(p->two->path));
2833                if (p->one->mode == 0)
2834                        len1 = snprintf(buffer, sizeof(buffer),
2835                                        "diff--gita/%.*sb/%.*s"
2836                                        "newfilemode%06o"
2837                                        "---/dev/null"
2838                                        "+++b/%.*s",
2839                                        len1, p->one->path,
2840                                        len2, p->two->path,
2841                                        p->two->mode,
2842                                        len2, p->two->path);
2843                else if (p->two->mode == 0)
2844                        len1 = snprintf(buffer, sizeof(buffer),
2845                                        "diff--gita/%.*sb/%.*s"
2846                                        "deletedfilemode%06o"
2847                                        "---a/%.*s"
2848                                        "+++/dev/null",
2849                                        len1, p->one->path,
2850                                        len2, p->two->path,
2851                                        p->one->mode,
2852                                        len1, p->one->path);
2853                else
2854                        len1 = snprintf(buffer, sizeof(buffer),
2855                                        "diff--gita/%.*sb/%.*s"
2856                                        "---a/%.*s"
2857                                        "+++b/%.*s",
2858                                        len1, p->one->path,
2859                                        len2, p->two->path,
2860                                        len1, p->one->path,
2861                                        len2, p->two->path);
2862                SHA1_Update(&ctx, buffer, len1);
2863
2864                xpp.flags = XDF_NEED_MINIMAL;
2865                xecfg.ctxlen = 3;
2866                xecfg.flags = XDL_EMIT_FUNCNAMES;
2867                ecb.outf = xdiff_outf;
2868                ecb.priv = &data;
2869                xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2870        }
2871
2872        SHA1_Final(sha1, &ctx);
2873        return 0;
2874}
2875
2876int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2877{
2878        struct diff_queue_struct *q = &diff_queued_diff;
2879        int i;
2880        int result = diff_get_patch_id(options, sha1);
2881
2882        for (i = 0; i < q->nr; i++)
2883                diff_free_filepair(q->queue[i]);
2884
2885        free(q->queue);
2886        q->queue = NULL;
2887        q->nr = q->alloc = 0;
2888
2889        return result;
2890}
2891
2892static int is_summary_empty(const struct diff_queue_struct *q)
2893{
2894        int i;
2895
2896        for (i = 0; i < q->nr; i++) {
2897                const struct diff_filepair *p = q->queue[i];
2898
2899                switch (p->status) {
2900                case DIFF_STATUS_DELETED:
2901                case DIFF_STATUS_ADDED:
2902                case DIFF_STATUS_COPIED:
2903                case DIFF_STATUS_RENAMED:
2904                        return 0;
2905                default:
2906                        if (p->score)
2907                                return 0;
2908                        if (p->one->mode && p->two->mode &&
2909                            p->one->mode != p->two->mode)
2910                                return 0;
2911                        break;
2912                }
2913        }
2914        return 1;
2915}
2916
2917void diff_flush(struct diff_options *options)
2918{
2919        struct diff_queue_struct *q = &diff_queued_diff;
2920        int i, output_format = options->output_format;
2921        int separator = 0;
2922
2923        /*
2924         * Order: raw, stat, summary, patch
2925         * or:    name/name-status/checkdiff (other bits clear)
2926         */
2927        if (!q->nr)
2928                goto free_queue;
2929
2930        if (output_format & (DIFF_FORMAT_RAW |
2931                             DIFF_FORMAT_NAME |
2932                             DIFF_FORMAT_NAME_STATUS |
2933                             DIFF_FORMAT_CHECKDIFF)) {
2934                for (i = 0; i < q->nr; i++) {
2935                        struct diff_filepair *p = q->queue[i];
2936                        if (check_pair_status(p))
2937                                flush_one_pair(p, options);
2938                }
2939                separator++;
2940        }
2941
2942        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2943                struct diffstat_t diffstat;
2944
2945                memset(&diffstat, 0, sizeof(struct diffstat_t));
2946                diffstat.xm.consume = diffstat_consume;
2947                for (i = 0; i < q->nr; i++) {
2948                        struct diff_filepair *p = q->queue[i];
2949                        if (check_pair_status(p))
2950                                diff_flush_stat(p, options, &diffstat);
2951                }
2952                if (output_format & DIFF_FORMAT_NUMSTAT)
2953                        show_numstat(&diffstat, options);
2954                if (output_format & DIFF_FORMAT_DIFFSTAT)
2955                        show_stats(&diffstat, options);
2956                else if (output_format & DIFF_FORMAT_SHORTSTAT)
2957                        show_shortstats(&diffstat);
2958                separator++;
2959        }
2960
2961        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2962                for (i = 0; i < q->nr; i++)
2963                        diff_summary(q->queue[i]);
2964                separator++;
2965        }
2966
2967        if (output_format & DIFF_FORMAT_PATCH) {
2968                if (separator) {
2969                        if (options->stat_sep) {
2970                                /* attach patch instead of inline */
2971                                fputs(options->stat_sep, stdout);
2972                        } else {
2973                                putchar(options->line_termination);
2974                        }
2975                }
2976
2977                for (i = 0; i < q->nr; i++) {
2978                        struct diff_filepair *p = q->queue[i];
2979                        if (check_pair_status(p))
2980                                diff_flush_patch(p, options);
2981                }
2982        }
2983
2984        if (output_format & DIFF_FORMAT_CALLBACK)
2985                options->format_callback(q, options, options->format_callback_data);
2986
2987        for (i = 0; i < q->nr; i++)
2988                diff_free_filepair(q->queue[i]);
2989free_queue:
2990        free(q->queue);
2991        q->queue = NULL;
2992        q->nr = q->alloc = 0;
2993}
2994
2995static void diffcore_apply_filter(const char *filter)
2996{
2997        int i;
2998        struct diff_queue_struct *q = &diff_queued_diff;
2999        struct diff_queue_struct outq;
3000        outq.queue = NULL;
3001        outq.nr = outq.alloc = 0;
3002
3003        if (!filter)
3004                return;
3005
3006        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3007                int found;
3008                for (i = found = 0; !found && i < q->nr; i++) {
3009                        struct diff_filepair *p = q->queue[i];
3010                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3011                             ((p->score &&
3012                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3013                              (!p->score &&
3014                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3015                            ((p->status != DIFF_STATUS_MODIFIED) &&
3016                             strchr(filter, p->status)))
3017                                found++;
3018                }
3019                if (found)
3020                        return;
3021
3022                /* otherwise we will clear the whole queue
3023                 * by copying the empty outq at the end of this
3024                 * function, but first clear the current entries
3025                 * in the queue.
3026                 */
3027                for (i = 0; i < q->nr; i++)
3028                        diff_free_filepair(q->queue[i]);
3029        }
3030        else {
3031                /* Only the matching ones */
3032                for (i = 0; i < q->nr; i++) {
3033                        struct diff_filepair *p = q->queue[i];
3034
3035                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3036                             ((p->score &&
3037                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3038                              (!p->score &&
3039                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3040                            ((p->status != DIFF_STATUS_MODIFIED) &&
3041                             strchr(filter, p->status)))
3042                                diff_q(&outq, p);
3043                        else
3044                                diff_free_filepair(p);
3045                }
3046        }
3047        free(q->queue);
3048        *q = outq;
3049}
3050
3051/* Check whether two filespecs with the same mode and size are identical */
3052static int diff_filespec_is_identical(struct diff_filespec *one,
3053                                      struct diff_filespec *two)
3054{
3055        if (S_ISGITLINK(one->mode)) {
3056                diff_fill_sha1_info(one);
3057                diff_fill_sha1_info(two);
3058                return !hashcmp(one->sha1, two->sha1);
3059        }
3060        if (diff_populate_filespec(one, 0))
3061                return 0;
3062        if (diff_populate_filespec(two, 0))
3063                return 0;
3064        return !memcmp(one->data, two->data, one->size);
3065}
3066
3067static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3068{
3069        int i;
3070        struct diff_queue_struct *q = &diff_queued_diff;
3071        struct diff_queue_struct outq;
3072        outq.queue = NULL;
3073        outq.nr = outq.alloc = 0;
3074
3075        for (i = 0; i < q->nr; i++) {
3076                struct diff_filepair *p = q->queue[i];
3077
3078                /*
3079                 * 1. Entries that come from stat info dirtyness
3080                 *    always have both sides (iow, not create/delete),
3081                 *    one side of the object name is unknown, with
3082                 *    the same mode and size.  Keep the ones that
3083                 *    do not match these criteria.  They have real
3084                 *    differences.
3085                 *
3086                 * 2. At this point, the file is known to be modified,
3087                 *    with the same mode and size, and the object
3088                 *    name of one side is unknown.  Need to inspect
3089                 *    the identical contents.
3090                 */
3091                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3092                    !DIFF_FILE_VALID(p->two) ||
3093                    (p->one->sha1_valid && p->two->sha1_valid) ||
3094                    (p->one->mode != p->two->mode) ||
3095                    diff_populate_filespec(p->one, 1) ||
3096                    diff_populate_filespec(p->two, 1) ||
3097                    (p->one->size != p->two->size) ||
3098                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3099                        diff_q(&outq, p);
3100                else {
3101                        /*
3102                         * The caller can subtract 1 from skip_stat_unmatch
3103                         * to determine how many paths were dirty only
3104                         * due to stat info mismatch.
3105                         */
3106                        if (!diffopt->no_index)
3107                                diffopt->skip_stat_unmatch++;
3108                        diff_free_filepair(p);
3109                }
3110        }
3111        free(q->queue);
3112        *q = outq;
3113}
3114
3115void diffcore_std(struct diff_options *options)
3116{
3117        if (options->quiet)
3118                return;
3119
3120        if (options->skip_stat_unmatch && !options->find_copies_harder)
3121                diffcore_skip_stat_unmatch(options);
3122        if (options->break_opt != -1)
3123                diffcore_break(options->break_opt);
3124        if (options->detect_rename)
3125                diffcore_rename(options);
3126        if (options->break_opt != -1)
3127                diffcore_merge_broken();
3128        if (options->pickaxe)
3129                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3130        if (options->orderfile)
3131                diffcore_order(options->orderfile);
3132        diff_resolve_rename_copy();
3133        diffcore_apply_filter(options->filter);
3134
3135        options->has_changes = !!diff_queued_diff.nr;
3136}
3137
3138
3139void diff_addremove(struct diff_options *options,
3140                    int addremove, unsigned mode,
3141                    const unsigned char *sha1,
3142                    const char *base, const char *path)
3143{
3144        char concatpath[PATH_MAX];
3145        struct diff_filespec *one, *two;
3146
3147        /* This may look odd, but it is a preparation for
3148         * feeding "there are unchanged files which should
3149         * not produce diffs, but when you are doing copy
3150         * detection you would need them, so here they are"
3151         * entries to the diff-core.  They will be prefixed
3152         * with something like '=' or '*' (I haven't decided
3153         * which but should not make any difference).
3154         * Feeding the same new and old to diff_change()
3155         * also has the same effect.
3156         * Before the final output happens, they are pruned after
3157         * merged into rename/copy pairs as appropriate.
3158         */
3159        if (options->reverse_diff)
3160                addremove = (addremove == '+' ? '-' :
3161                             addremove == '-' ? '+' : addremove);
3162
3163        if (!path) path = "";
3164        sprintf(concatpath, "%s%s", base, path);
3165        one = alloc_filespec(concatpath);
3166        two = alloc_filespec(concatpath);
3167
3168        if (addremove != '+')
3169                fill_filespec(one, sha1, mode);
3170        if (addremove != '-')
3171                fill_filespec(two, sha1, mode);
3172
3173        diff_queue(&diff_queued_diff, one, two);
3174        options->has_changes = 1;
3175}
3176
3177void diff_change(struct diff_options *options,
3178                 unsigned old_mode, unsigned new_mode,
3179                 const unsigned char *old_sha1,
3180                 const unsigned char *new_sha1,
3181                 const char *base, const char *path)
3182{
3183        char concatpath[PATH_MAX];
3184        struct diff_filespec *one, *two;
3185
3186        if (options->reverse_diff) {
3187                unsigned tmp;
3188                const unsigned char *tmp_c;
3189                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3190                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3191        }
3192        if (!path) path = "";
3193        sprintf(concatpath, "%s%s", base, path);
3194        one = alloc_filespec(concatpath);
3195        two = alloc_filespec(concatpath);
3196        fill_filespec(one, old_sha1, old_mode);
3197        fill_filespec(two, new_sha1, new_mode);
3198
3199        diff_queue(&diff_queued_diff, one, two);
3200        options->has_changes = 1;
3201}
3202
3203void diff_unmerge(struct diff_options *options,
3204                  const char *path,
3205                  unsigned mode, const unsigned char *sha1)
3206{
3207        struct diff_filespec *one, *two;
3208        one = alloc_filespec(path);
3209        two = alloc_filespec(path);
3210        fill_filespec(one, sha1, mode);
3211        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3212}