tar-tree.con commit archimport: safer log file parsing (6df896b)
   1/*
   2 * Copyright (c) 2005 Rene Scharfe
   3 */
   4#include <time.h>
   5#include "cache.h"
   6
   7#define RECORDSIZE      (512)
   8#define BLOCKSIZE       (RECORDSIZE * 20)
   9
  10#define TYPEFLAG_AUTO           '\0'
  11#define TYPEFLAG_REG            '0'
  12#define TYPEFLAG_LNK            '2'
  13#define TYPEFLAG_DIR            '5'
  14#define TYPEFLAG_GLOBAL_HEADER  'g'
  15#define TYPEFLAG_EXT_HEADER     'x'
  16
  17#define EXT_HEADER_PATH         1
  18#define EXT_HEADER_LINKPATH     2
  19
  20static const char tar_tree_usage[] = "git-tar-tree <key> [basedir]";
  21
  22static char block[BLOCKSIZE];
  23static unsigned long offset;
  24
  25static const char *basedir;
  26static time_t archive_time;
  27
  28struct path_prefix {
  29        struct path_prefix *prev;
  30        const char *name;
  31};
  32
  33/* tries hard to write, either succeeds or dies in the attempt */
  34static void reliable_write(void *buf, unsigned long size)
  35{
  36        while (size > 0) {
  37                long ret = write(1, buf, size);
  38                if (ret < 0) {
  39                        if (errno == EAGAIN)
  40                                continue;
  41                        if (errno == EPIPE)
  42                                exit(0);
  43                        die("git-tar-tree: %s", strerror(errno));
  44                } else if (!ret) {
  45                        die("git-tar-tree: disk full?");
  46                }
  47                size -= ret;
  48                buf += ret;
  49        }
  50}
  51
  52/* writes out the whole block, but only if it is full */
  53static void write_if_needed(void)
  54{
  55        if (offset == BLOCKSIZE) {
  56                reliable_write(block, BLOCKSIZE);
  57                offset = 0;
  58        }
  59}
  60
  61/* acquire the next record from the buffer; user must call write_if_needed() */
  62static char *get_record(void)
  63{
  64        char *p = block + offset;
  65        memset(p, 0, RECORDSIZE);
  66        offset += RECORDSIZE;
  67        return p;
  68}
  69
  70/*
  71 * The end of tar archives is marked by 1024 nul bytes and after that
  72 * follows the rest of the block (if any).
  73 */
  74static void write_trailer(void)
  75{
  76        get_record();
  77        write_if_needed();
  78        get_record();
  79        write_if_needed();
  80        while (offset) {
  81                get_record();
  82                write_if_needed();
  83        }
  84}
  85
  86/*
  87 * queues up writes, so that all our write(2) calls write exactly one
  88 * full block; pads writes to RECORDSIZE
  89 */
  90static void write_blocked(void *buf, unsigned long size)
  91{
  92        unsigned long tail;
  93
  94        if (offset) {
  95                unsigned long chunk = BLOCKSIZE - offset;
  96                if (size < chunk)
  97                        chunk = size;
  98                memcpy(block + offset, buf, chunk);
  99                size -= chunk;
 100                offset += chunk;
 101                buf += chunk;
 102                write_if_needed();
 103        }
 104        while (size >= BLOCKSIZE) {
 105                reliable_write(buf, BLOCKSIZE);
 106                size -= BLOCKSIZE;
 107                buf += BLOCKSIZE;
 108        }
 109        if (size) {
 110                memcpy(block + offset, buf, size);
 111                buf += size;
 112                offset += size;
 113        }
 114        tail = offset % RECORDSIZE;
 115        if (tail)  {
 116                memset(block + offset, 0, RECORDSIZE - tail);
 117                offset += RECORDSIZE - tail;
 118        }
 119        write_if_needed();
 120}
 121
 122static void append_string(char **p, const char *s)
 123{
 124        unsigned int len = strlen(s);
 125        memcpy(*p, s, len);
 126        *p += len;
 127}
 128
 129static void append_char(char **p, char c)
 130{
 131        **p = c;
 132        *p += 1;
 133}
 134
 135static void append_path_prefix(char **buffer, struct path_prefix *prefix)
 136{
 137        if (!prefix)
 138                return;
 139        append_path_prefix(buffer, prefix->prev);
 140        append_string(buffer, prefix->name);
 141        append_char(buffer, '/');
 142}
 143
 144static unsigned int path_prefix_len(struct path_prefix *prefix)
 145{
 146        if (!prefix)
 147                return 0;
 148        return path_prefix_len(prefix->prev) + strlen(prefix->name) + 1;
 149}
 150
 151static void append_path(char **p, int is_dir, const char *basepath,
 152                        struct path_prefix *prefix, const char *path)
 153{
 154        if (basepath) {
 155                append_string(p, basepath);
 156                append_char(p, '/');
 157        }
 158        append_path_prefix(p, prefix);
 159        append_string(p, path);
 160        if (is_dir)
 161                append_char(p, '/');
 162}
 163
 164static unsigned int path_len(int is_dir, const char *basepath,
 165                             struct path_prefix *prefix, const char *path)
 166{
 167        unsigned int len = 0;
 168        if (basepath)
 169                len += strlen(basepath) + 1;
 170        len += path_prefix_len(prefix) + strlen(path);
 171        if (is_dir)
 172                len++;
 173        return len;
 174}
 175
 176static void append_extended_header_prefix(char **p, unsigned int size,
 177                                          const char *keyword)
 178{
 179        int len = sprintf(*p, "%u %s=", size, keyword);
 180        *p += len;
 181}
 182
 183static unsigned int extended_header_len(const char *keyword,
 184                                        unsigned int valuelen)
 185{
 186        /* "%u %s=%s\n" */
 187        unsigned int len = 1 + 1 + strlen(keyword) + 1 + valuelen + 1;
 188        if (len > 9)
 189                len++;
 190        if (len > 99)
 191                len++;
 192        return len;
 193}
 194
 195static void append_extended_header(char **p, const char *keyword,
 196                                   const char *value, unsigned int len)
 197{
 198        unsigned int size = extended_header_len(keyword, len);
 199        append_extended_header_prefix(p, size, keyword);
 200        memcpy(*p, value, len);
 201        *p += len;
 202        append_char(p, '\n');
 203}
 204
 205static void write_header(const unsigned char *, char, const char *, struct path_prefix *,
 206                         const char *, unsigned int, void *, unsigned long);
 207
 208/* stores a pax extended header directly in the block buffer */
 209static void write_extended_header(const char *headerfilename, int is_dir,
 210                                  unsigned int flags, const char *basepath,
 211                                  struct path_prefix *prefix,
 212                                  const char *path, unsigned int namelen,
 213                                  void *content, unsigned int contentsize)
 214{
 215        char *buffer, *p;
 216        unsigned int pathlen, size, linkpathlen = 0;
 217
 218        size = pathlen = extended_header_len("path", namelen);
 219        if (flags & EXT_HEADER_LINKPATH) {
 220                linkpathlen = extended_header_len("linkpath", contentsize);
 221                size += linkpathlen;
 222        }
 223        write_header(NULL, TYPEFLAG_EXT_HEADER, NULL, NULL, headerfilename,
 224                     0100600, NULL, size);
 225
 226        buffer = p = malloc(size);
 227        if (!buffer)
 228                die("git-tar-tree: %s", strerror(errno));
 229        append_extended_header_prefix(&p, pathlen, "path");
 230        append_path(&p, is_dir, basepath, prefix, path);
 231        append_char(&p, '\n');
 232        if (flags & EXT_HEADER_LINKPATH)
 233                append_extended_header(&p, "linkpath", content, contentsize);
 234        write_blocked(buffer, size);
 235        free(buffer);
 236}
 237
 238static void write_global_extended_header(const unsigned char *sha1)
 239{
 240        char *p;
 241        unsigned int size;
 242
 243        size = extended_header_len("comment", 40);
 244        write_header(NULL, TYPEFLAG_GLOBAL_HEADER, NULL, NULL,
 245                     "pax_global_header", 0100600, NULL, size);
 246
 247        p = get_record();
 248        append_extended_header(&p, "comment", sha1_to_hex(sha1), 40);
 249        write_if_needed();
 250}
 251
 252/* stores a ustar header directly in the block buffer */
 253static void write_header(const unsigned char *sha1, char typeflag, const char *basepath,
 254                         struct path_prefix *prefix, const char *path,
 255                         unsigned int mode, void *buffer, unsigned long size)
 256{
 257        unsigned int namelen; 
 258        char *header = NULL;
 259        unsigned int checksum = 0;
 260        int i;
 261        unsigned int ext_header = 0;
 262
 263        if (typeflag == TYPEFLAG_AUTO) {
 264                if (S_ISDIR(mode))
 265                        typeflag = TYPEFLAG_DIR;
 266                else if (S_ISLNK(mode))
 267                        typeflag = TYPEFLAG_LNK;
 268                else
 269                        typeflag = TYPEFLAG_REG;
 270        }
 271
 272        namelen = path_len(S_ISDIR(mode), basepath, prefix, path);
 273        if (namelen > 100)
 274                ext_header |= EXT_HEADER_PATH;
 275        if (typeflag == TYPEFLAG_LNK && size > 100)
 276                ext_header |= EXT_HEADER_LINKPATH;
 277
 278        /* the extended header must be written before the normal one */
 279        if (ext_header) {
 280                char headerfilename[51];
 281                sprintf(headerfilename, "%s.paxheader", sha1_to_hex(sha1));
 282                write_extended_header(headerfilename, S_ISDIR(mode),
 283                                      ext_header, basepath, prefix, path,
 284                                      namelen, buffer, size);
 285        }
 286
 287        header = get_record();
 288
 289        if (ext_header) {
 290                sprintf(header, "%s.data", sha1_to_hex(sha1));
 291        } else {
 292                char *p = header;
 293                append_path(&p, S_ISDIR(mode), basepath, prefix, path);
 294        }
 295
 296        if (typeflag == TYPEFLAG_LNK) {
 297                if (ext_header & EXT_HEADER_LINKPATH) {
 298                        sprintf(&header[157], "see %s.paxheader",
 299                                sha1_to_hex(sha1));
 300                } else {
 301                        if (buffer)
 302                                strncpy(&header[157], buffer, size);
 303                }
 304        }
 305
 306        if (S_ISDIR(mode))
 307                mode |= 0755;   /* GIT doesn't store permissions of dirs */
 308        if (S_ISLNK(mode))
 309                mode |= 0777;   /* ... nor of symlinks */
 310        sprintf(&header[100], "%07o", mode & 07777);
 311
 312        /* XXX: should we provide more meaningful info here? */
 313        sprintf(&header[108], "%07o", 0);       /* uid */
 314        sprintf(&header[116], "%07o", 0);       /* gid */
 315        strncpy(&header[265], "git", 31);       /* uname */
 316        strncpy(&header[297], "git", 31);       /* gname */
 317
 318        if (S_ISDIR(mode) || S_ISLNK(mode))
 319                size = 0;
 320        sprintf(&header[124], "%011lo", size);
 321        sprintf(&header[136], "%011lo", archive_time);
 322
 323        header[156] = typeflag;
 324
 325        memcpy(&header[257], "ustar", 6);
 326        memcpy(&header[263], "00", 2);
 327
 328        sprintf(&header[329], "%07o", 0);       /* devmajor */
 329        sprintf(&header[337], "%07o", 0);       /* devminor */
 330
 331        memset(&header[148], ' ', 8);
 332        for (i = 0; i < RECORDSIZE; i++)
 333                checksum += header[i];
 334        sprintf(&header[148], "%07o", checksum & 0x1fffff);
 335
 336        write_if_needed();
 337}
 338
 339static void traverse_tree(void *buffer, unsigned long size,
 340                          struct path_prefix *prefix)
 341{
 342        struct path_prefix this_prefix;
 343        this_prefix.prev = prefix;
 344
 345        while (size) {
 346                int namelen = strlen(buffer)+1;
 347                void *eltbuf;
 348                char elttype[20];
 349                unsigned long eltsize;
 350                unsigned char *sha1 = buffer + namelen;
 351                char *path = strchr(buffer, ' ') + 1;
 352                unsigned int mode;
 353
 354                if (size < namelen + 20 || sscanf(buffer, "%o", &mode) != 1)
 355                        die("corrupt 'tree' file");
 356                if (S_ISDIR(mode) || S_ISREG(mode))
 357                        mode |= (mode & 0100) ? 0777 : 0666;
 358                buffer = sha1 + 20;
 359                size -= namelen + 20;
 360
 361                eltbuf = read_sha1_file(sha1, elttype, &eltsize);
 362                if (!eltbuf)
 363                        die("cannot read %s", sha1_to_hex(sha1));
 364                write_header(sha1, TYPEFLAG_AUTO, basedir, prefix, path,
 365                             mode, eltbuf, eltsize);
 366                if (!strcmp(elttype, "tree")) {
 367                        this_prefix.name = path;
 368                        traverse_tree(eltbuf, eltsize, &this_prefix);
 369                } else if (!strcmp(elttype, "blob") && !S_ISLNK(mode)) {
 370                        write_blocked(eltbuf, eltsize);
 371                }
 372                free(eltbuf);
 373        }
 374}
 375
 376/* get commit time from committer line of commit object */
 377static time_t commit_time(void * buffer, unsigned long size)
 378{
 379        time_t result = 0;
 380        char *p = buffer;
 381
 382        while (size > 0) {
 383                char *endp = memchr(p, '\n', size);
 384                if (!endp || endp == p)
 385                        break;
 386                *endp = '\0';
 387                if (endp - p > 10 && !memcmp(p, "committer ", 10)) {
 388                        char *nump = strrchr(p, '>');
 389                        if (!nump)
 390                                break;
 391                        nump++;
 392                        result = strtoul(nump, &endp, 10);
 393                        if (*endp != ' ')
 394                                result = 0;
 395                        break;
 396                }
 397                size -= endp - p - 1;
 398                p = endp + 1;
 399        }
 400        return result;
 401}
 402
 403int main(int argc, char **argv)
 404{
 405        unsigned char sha1[20];
 406        unsigned char commit_sha1[20];
 407        void *buffer;
 408        unsigned long size;
 409
 410        setup_git_directory();
 411
 412        switch (argc) {
 413        case 3:
 414                basedir = argv[2];
 415                /* FALLTHROUGH */
 416        case 2:
 417                if (get_sha1(argv[1], sha1) < 0)
 418                        usage(tar_tree_usage);
 419                break;
 420        default:
 421                usage(tar_tree_usage);
 422        }
 423
 424        buffer = read_object_with_reference(sha1, "commit", &size, commit_sha1);
 425        if (buffer) {
 426                write_global_extended_header(commit_sha1);
 427                archive_time = commit_time(buffer, size);
 428                free(buffer);
 429        }
 430        buffer = read_object_with_reference(sha1, "tree", &size, NULL);
 431        if (!buffer)
 432                die("not a reference to a tag, commit or tree object: %s",
 433                    sha1_to_hex(sha1));
 434        if (!archive_time)
 435                archive_time = time(NULL);
 436        if (basedir)
 437                write_header((unsigned char *)"0", TYPEFLAG_DIR, NULL, NULL,
 438                        basedir, 040755, NULL, 0);
 439        traverse_tree(buffer, size, NULL);
 440        free(buffer);
 441        write_trailer();
 442        return 0;
 443}