91a71c55c6bb720c7fbe3d1d2bd393d5d22f764f
   1#include "cache.h"
   2#include "object.h"
   3#include "delta.h"
   4
   5static int dry_run;
   6static int nr_entries;
   7static const char *base_name;
   8static const char unpack_usage[] = "git-unpack-objects basename";
   9
  10struct pack_entry {
  11        unsigned int offset; /* network byte order */
  12        unsigned char sha1[20];
  13};
  14
  15static void *pack_base;
  16static unsigned long pack_size;
  17static void *index_base;
  18static unsigned long index_size;
  19
  20static struct pack_entry **pack_list;
  21
  22static void *map_file(const char *suffix, unsigned long *sizep)
  23{
  24        static char pathname[PATH_MAX];
  25        unsigned long len;
  26        int fd;
  27        struct stat st;
  28        void *map;
  29
  30        len = snprintf(pathname, PATH_MAX, "%s.%s", base_name, suffix);
  31        if (len >= PATH_MAX)
  32                die("bad pack base-name");
  33        fd = open(pathname, O_RDONLY);
  34        if (fd < 0 || fstat(fd, &st))
  35                die("unable to open '%s'", pathname);
  36        len = st.st_size;
  37        if (!len)
  38                die("bad pack file '%s'", pathname);
  39        map = mmap(NULL, len, PROT_READ, MAP_PRIVATE, fd, 0);
  40        if (-1 == (int)(long)map)
  41                die("unable to mmap '%s'", pathname);
  42        close(fd);
  43        *sizep = len;
  44        return map;
  45}
  46
  47static int sort_by_offset(const void *_a, const void *_b)
  48{
  49        struct pack_entry *a = *(struct pack_entry **)_a;
  50        struct pack_entry *b = *(struct pack_entry **)_b;
  51        unsigned int o1, o2;
  52
  53        o1 = ntohl(a->offset);
  54        o2 = ntohl(b->offset);
  55        return o1 < o2 ? -1 : 1;
  56}
  57
  58static int check_index(void)
  59{
  60        unsigned int *array = index_base;
  61        unsigned int nr;
  62        int i;
  63
  64        if (index_size < 4*256 + 20)
  65                return error("index file too small");
  66        nr = 0;
  67        for (i = 0; i < 256; i++) {
  68                unsigned int n = ntohl(array[i]);
  69                if (n < nr)
  70                        return error("non-monotonic index");
  71                nr = n;
  72        }
  73        /*
  74         * Total size:
  75         *  - 256 index entries 4 bytes each
  76         *  - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
  77         *  - 20-byte SHA1 file checksum
  78         */
  79        if (index_size != 4*256 + nr * 24 + 20)
  80                return error("wrong index file size");
  81
  82        nr_entries = nr;
  83        pack_list = xmalloc(nr * sizeof(struct pack_entry *));
  84        for (i = 0; i < nr; i++)
  85                pack_list[i] = index_base + 4*256 + i*24;
  86
  87        qsort(pack_list, nr, sizeof(*pack_list), sort_by_offset);
  88
  89        printf("%d entries\n", nr);
  90        return 0;
  91}
  92
  93static int unpack_non_delta_entry(struct pack_entry *entry,
  94                                  int kind,
  95                                  unsigned char *data,
  96                                  unsigned long size,
  97                                  unsigned long left)
  98{
  99        int st;
 100        z_stream stream;
 101        char *buffer;
 102        unsigned char sha1[20];
 103        char *type_s;
 104
 105        printf("%s %c %lu\n", sha1_to_hex(entry->sha1), kind, size);
 106        if (dry_run)
 107                return 0;
 108
 109        buffer = xmalloc(size + 1);
 110        buffer[size] = 0;
 111        memset(&stream, 0, sizeof(stream));
 112        stream.next_in = data;
 113        stream.avail_in = left;
 114        stream.next_out = buffer;
 115        stream.avail_out = size;
 116
 117        inflateInit(&stream);
 118        st = inflate(&stream, Z_FINISH);
 119        inflateEnd(&stream);
 120        if ((st != Z_STREAM_END) || stream.total_out != size)
 121                goto err_finish;
 122        switch (kind) {
 123        case 'C': type_s = "commit"; break;
 124        case 'T': type_s = "tree"; break;
 125        case 'B': type_s = "blob"; break;
 126        default: goto err_finish;
 127        }
 128        if (write_sha1_file(buffer, size, type_s, sha1) < 0)
 129                die("failed to write %s (%s)",
 130                    sha1_to_hex(entry->sha1), type_s);
 131        printf("%s %s\n", sha1_to_hex(sha1), type_s);
 132        if (memcmp(sha1, entry->sha1, 20))
 133                die("resulting %s have wrong SHA1", type_s);
 134
 135 finish:
 136        st = 0;
 137        free(buffer);
 138        return st;
 139 err_finish:
 140        st = -1;
 141        goto finish;
 142}
 143
 144static int find_pack_entry(unsigned char *sha1, struct pack_entry **ent)
 145{
 146        int *level1_ofs = index_base;
 147        int hi = ntohl(level1_ofs[*sha1]);
 148        int lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
 149        void *index = index_base + 4*256;
 150
 151        do {
 152                int mi = (lo + hi) / 2;
 153                int cmp = memcmp(index + 24 * mi + 4, sha1, 20);
 154                if (!cmp) {
 155                        *ent = index + 24 * mi;
 156                        return 1;
 157                }
 158                if (cmp > 0)
 159                        hi = mi;
 160                else
 161                        lo = mi+1;
 162        } while (lo < hi);
 163        return 0;
 164}
 165
 166/* forward declaration for a mutually recursive function */
 167static void unpack_entry(struct pack_entry *);
 168
 169static int unpack_delta_entry(struct pack_entry *entry,
 170                              unsigned char *base_sha1,
 171                              unsigned long delta_size,
 172                              unsigned long left)
 173{
 174        void *data, *delta_data, *result, *base;
 175        unsigned long data_size, result_size, base_size;
 176        z_stream stream;
 177        int st;
 178        char type[20];
 179        unsigned char sha1[20];
 180
 181        if (left < 20)
 182                die("truncated pack file");
 183        data = base_sha1 + 20;
 184        data_size = left - 20;
 185        printf("%s D %lu", sha1_to_hex(entry->sha1), delta_size);
 186        printf(" %s\n", sha1_to_hex(base_sha1));
 187
 188        if (dry_run)
 189                return 0;
 190
 191        /* pack+5 is the base sha1, unless we have it, we need to
 192         * unpack it first.
 193         */
 194        if (!has_sha1_file(base_sha1)) {
 195                struct pack_entry *base;
 196                if (!find_pack_entry(base_sha1, &base))
 197                        die("cannot find delta-pack base object");
 198                unpack_entry(base);
 199        }
 200        delta_data = xmalloc(delta_size);
 201
 202        memset(&stream, 0, sizeof(stream));
 203
 204        stream.next_in = data;
 205        stream.avail_in = data_size;
 206        stream.next_out = delta_data;
 207        stream.avail_out = delta_size;
 208
 209        inflateInit(&stream);
 210        st = inflate(&stream, Z_FINISH);
 211        inflateEnd(&stream);
 212        if ((st != Z_STREAM_END) || stream.total_out != delta_size)
 213                die("delta data unpack failed");
 214
 215        base = read_sha1_file(base_sha1, type, &base_size);
 216        if (!base)
 217                die("failed to read delta-pack base object %s", sha1_to_hex(base_sha1));
 218        result = patch_delta(base, base_size,
 219                             delta_data, delta_size,
 220                             &result_size);
 221        if (!result)
 222                die("failed to apply delta");
 223        free(delta_data);
 224
 225        if (write_sha1_file(result, result_size, type, sha1) < 0)
 226                die("failed to write %s (%s)",
 227                    sha1_to_hex(entry->sha1), type);
 228        free(result);
 229        printf("%s %s\n", sha1_to_hex(sha1), type);
 230        if (memcmp(sha1, entry->sha1, 20))
 231                die("resulting %s have wrong SHA1", type);
 232        return 0;
 233}
 234
 235static void unpack_entry(struct pack_entry *entry)
 236{
 237        unsigned long offset, size, left;
 238        unsigned char *pack;
 239
 240        /* Have we done this one already due to deltas based on it? */
 241        if (lookup_object(entry->sha1))
 242                return;
 243
 244        offset = ntohl(entry->offset);
 245        if (offset > pack_size - 5)
 246                die("object offset outside of pack file");
 247        pack = pack_base + offset;
 248        size = (pack[1] << 24) + (pack[2] << 16) + (pack[3] << 8) + pack[4];
 249        left = pack_size - offset - 5;
 250        switch (*pack) {
 251        case 'C': case 'T': case 'B':
 252                unpack_non_delta_entry(entry, *pack, pack+5, size, left);
 253                break;
 254        case 'D':
 255                unpack_delta_entry(entry, pack+5, size, left);
 256                break;
 257        default:
 258                die("corrupted pack file");
 259        }
 260}
 261
 262/*
 263 * We unpack from the end, older files first. Now, usually
 264 * there are deltas etc, so we'll not actually write the
 265 * objects in that order, but we might as well try..
 266 */
 267static void unpack_all(void)
 268{
 269        int i = nr_entries;
 270
 271        while (--i >= 0) {
 272                struct pack_entry *entry = pack_list[i];
 273                unpack_entry(entry);
 274        }
 275}
 276
 277int main(int argc, char **argv)
 278{
 279        int i;
 280
 281        for (i = 1 ; i < argc; i++) {
 282                const char *arg = argv[i];
 283
 284                if (*arg == '-') {
 285                        if (!strcmp(arg, "-n")) {
 286                                dry_run = 1;
 287                                continue;
 288                        }
 289                        usage(unpack_usage);
 290                }
 291                if (base_name)
 292                        usage(unpack_usage);
 293                base_name = arg;
 294        }
 295        if (!base_name)
 296                usage(unpack_usage);
 297        index_base = map_file("idx", &index_size);
 298        pack_base = map_file("pack", &pack_size);
 299        if (check_index() < 0)
 300                die("bad index file");
 301        unpack_all();
 302        return 0;
 303}