unpack-objects.c

   1 #include "cache.h"
   2 #include "object.h"
   3 #include "delta.h"
   4 #include "pack.h"
   5
   6 static int dry_run;
   7 static const char unpack_usage[] = "git-unpack-objects < pack-file";
   8
   9 /* We always read in 4kB chunks. */
  10 static unsigned char buffer[4096];
  11 static unsigned long offset, len, eof;
  12 static SHA_CTX ctx;
  13
  14 /*
  15  * Make sure at least "min" bytes are available in the buffer, and
  16  * return the pointer to the buffer.
  17  */
  18 static void * fill(int min)
  19 {
  20         if (min <= len)
  21                 return buffer + offset;
  22         if (eof)
  23                 die("unable to fill input");
  24         if (min > sizeof(buffer))
  25                 die("cannot fill %d bytes", min);
  26         if (offset) {
  27                 SHA1_Update(&ctx, buffer, offset);
  28                 memcpy(buffer, buffer + offset, len);
  29                 offset = 0;
  30         }
  31         do {
  32                 int ret = read(0, buffer + len, sizeof(buffer) - len);
  33                 if (ret <= 0) {
  34                         if (!ret)
  35                                 die("early EOF");
  36                         if (errno == EAGAIN || errno == EINTR)
  37                                 continue;
  38                         die("read error on input: %s", strerror(errno));
  39                 }
  40                 len += ret;
  41         } while (len < min);
  42         return buffer;
  43 }
  44
  45 static void use(int bytes)
  46 {
  47         if (bytes > len)
  48                 die("used more bytes than were available");
  49         len -= bytes;
  50         offset += bytes;
  51 }
  52
  53 static void *get_data(unsigned long size)
  54 {
  55         z_stream stream;
  56         void *buf = xmalloc(size);
  57
  58         memset(&stream, 0, sizeof(stream));
  59
  60         stream.next_out = buf;
  61         stream.avail_out = size;
  62         stream.next_in = fill(1);
  63         stream.avail_in = len;
  64         inflateInit(&stream);
  65
  66         for (;;) {
  67                 int ret = inflate(&stream, 0);
  68                 use(len - stream.avail_in);
  69                 if (stream.total_out == size && ret == Z_STREAM_END)
  70                         break;
  71                 if (ret != Z_OK)
  72                         die("inflate returned %d\n", ret);
  73                 stream.next_in = fill(1);
  74                 stream.avail_in = len;
  75         }
  76         return buf;
  77 }
  78
  79 struct delta_info {
  80         unsigned char base_sha1[20];
  81         unsigned long size;
  82         void *delta;
  83         struct delta_info *next;
  84 };
  85
  86 static struct delta_info *delta_list;
  87
  88 static void add_delta_to_list(unsigned char *base_sha1, void *delta, unsigned long size)
  89 {
  90         struct delta_info *info = xmalloc(sizeof(*info));
  91
  92         memcpy(info->base_sha1, base_sha1, 20);
  93         info->size = size;
  94         info->delta = delta;
  95         info->next = delta_list;
  96         delta_list = info;
  97 }
  98
  99 static void added_object(unsigned char *sha1, const char *type, void *data, unsigned long size);
 100
 101 static void write_object(void *buf, unsigned long size, const char *type)
 102 {
 103         unsigned char sha1[20];
 104         if (write_sha1_file(buf, size, type, sha1) < 0)
 105                 die("failed to write object");
 106         added_object(sha1, type, buf, size);
 107 }
 108
 109 static int resolve_delta(const char *type,
 110         void *base, unsigned long base_size,
 111         void *delta, unsigned long delta_size)
 112 {
 113         void *result;
 114         unsigned long result_size;
 115
 116         result = patch_delta(base, base_size,
 117                              delta, delta_size,
 118                              &result_size);
 119         if (!result)
 120                 die("failed to apply delta");
 121         free(delta);
 122         write_object(result, result_size, type);
 123         free(result);
 124         return 0;
 125 }
 126
 127 static void added_object(unsigned char *sha1, const char *type, void *data, unsigned long size)
 128 {
 129         struct delta_info **p = &delta_list;
 130         struct delta_info *info;
 131
 132         while ((info = *p) != NULL) {
 133                 if (!memcmp(info->base_sha1, sha1, 20)) {
 134                         *p = info->next;
 135                         p = &delta_list;
 136                         resolve_delta(type, data, size, info->delta, info->size);
 137                         free(info);
 138                         continue;
 139                 }
 140                 p = &info->next;
 141         }
 142 }
 143
 144 static int unpack_non_delta_entry(enum object_type kind, unsigned long size)
 145 {
 146         void *buf = get_data(size);
 147         const char *type;
 148
 149         switch (kind) {
 150         case OBJ_COMMIT: type = "commit"; break;
 151         case OBJ_TREE:   type = "tree"; break;
 152         case OBJ_BLOB:   type = "blob"; break;
 153         case OBJ_TAG:    type = "tag"; break;
 154         default: die("bad type %d", kind);
 155         }
 156         if (!dry_run)
 157                 write_object(buf, size, type);
 158         free(buf);
 159         return 0;
 160 }
 161
 162 static int unpack_delta_entry(unsigned long delta_size)
 163 {
 164         void *delta_data, *base;
 165         unsigned long base_size;
 166         char type[20];
 167         unsigned char base_sha1[20];
 168
 169         memcpy(base_sha1, fill(20), 20);
 170         use(20);
 171
 172         delta_data = get_data(delta_size);
 173         if (dry_run) {
 174                 free(delta_data);
 175                 return 0;
 176         }
 177
 178         if (!has_sha1_file(base_sha1)) {
 179                 add_delta_to_list(base_sha1, delta_data, delta_size);
 180                 return 0;
 181         }
 182         base = read_sha1_file(base_sha1, type, &base_size);
 183         if (!base)
 184                 die("failed to read delta-pack base object %s", sha1_to_hex(base_sha1));
 185         return resolve_delta(type, base, base_size, delta_data, delta_size);
 186 }
 187
 188 static void unpack_one(void)
 189 {
 190         unsigned shift;
 191         unsigned char *pack, c;
 192         unsigned long size;
 193         enum object_type type;
 194
 195         pack = fill(1);
 196         c = *pack;
 197         use(1);
 198         type = (c >> 4) & 7;
 199         size = (c & 15);
 200         shift = 4;
 201         while (c & 0x80) {
 202                 pack = fill(1);
 203                 c = *pack++;
 204                 use(1);
 205                 size += (c & 0x7f) << shift;
 206                 shift += 7;
 207         }
 208         switch (type) {
 209         case OBJ_COMMIT:
 210         case OBJ_TREE:
 211         case OBJ_BLOB:
 212         case OBJ_TAG:
 213                 unpack_non_delta_entry(type, size);
 214                 return;
 215         case OBJ_DELTA:
 216                 unpack_delta_entry(size);
 217                 return;
 218         default:
 219                 die("bad object type %d", type);
 220         }
 221 }
 222
 223 /*
 224  * We unpack from the end, older files first. Now, usually
 225  * there are deltas etc, so we'll not actually write the
 226  * objects in that order, but we might as well try..
 227  */
 228 static void unpack_all(void)
 229 {
 230         int i;
 231         struct pack_header *hdr = fill(sizeof(struct pack_header));
 232         unsigned version = ntohl(hdr->hdr_version);
 233         unsigned nr_objects = ntohl(hdr->hdr_entries);
 234
 235         if (ntohl(hdr->hdr_signature) != PACK_SIGNATURE)
 236                 die("bad pack file");
 237         if (version != PACK_VERSION)
 238                 die("unable to handle pack file version %d", version);
 239         fprintf(stderr, "Unpacking %d objects\n", nr_objects);
 240
 241         use(sizeof(struct pack_header));
 242         for (i = 0; i < nr_objects; i++)
 243                 unpack_one();
 244         if (delta_list)
 245                 die("unresolved deltas left after unpacking");
 246 }
 247
 248 int main(int argc, char **argv)
 249 {
 250         int i;
 251         unsigned char sha1[20];
 252
 253         for (i = 1 ; i < argc; i++) {
 254                 const char *arg = argv[i];
 255
 256                 if (*arg == '-') {
 257                         if (!strcmp(arg, "-n")) {
 258                                 dry_run = 1;
 259                                 continue;
 260                         }
 261                         usage(unpack_usage);
 262                 }
 263
 264                 /* We don't take any non-flag arguments now.. Maybe some day */
 265                 usage(unpack_usage);
 266         }
 267         SHA1_Init(&ctx);
 268         unpack_all();
 269         SHA1_Update(&ctx, buffer, offset);
 270         SHA1_Final(sha1, &ctx);
 271         if (memcmp(fill(20), sha1, 20))
 272                 die("final sha1 did not match");
 273         use(20);
 274
 275         /* Write the last part of the buffer to stdout */
 276         while (len) {
 277                 int ret = write(1, buffer + offset, len);
 278                 if (!ret)
 279                         break;
 280                 if (ret < 0) {
 281                         if (errno == EAGAIN || errno == EINTR)
 282                                 continue;
 283                         break;
 284                 }
 285                 len -= ret;
 286                 offset += ret;
 287         }
 288
 289         /* All done */
 290         return 0;
 291 }