65bd335aa26f71b2df82ae490e354736a3a73992
   1/*
   2 * Parse and rearrange a svnadmin dump.
   3 * Create the dump with:
   4 * svnadmin dump --incremental -r<startrev>:<endrev> <repository> >outfile
   5 *
   6 * Licensed under a two-clause BSD-style license.
   7 * See LICENSE for details.
   8 */
   9
  10#include "cache.h"
  11#include "repo_tree.h"
  12#include "fast_export.h"
  13#include "line_buffer.h"
  14#include "obj_pool.h"
  15#include "string_pool.h"
  16
  17#define NODEACT_REPLACE 4
  18#define NODEACT_DELETE 3
  19#define NODEACT_ADD 2
  20#define NODEACT_CHANGE 1
  21#define NODEACT_UNKNOWN 0
  22
  23#define DUMP_CTX 0
  24#define REV_CTX  1
  25#define NODE_CTX 2
  26
  27#define LENGTH_UNKNOWN (~0)
  28#define DATE_RFC2822_LEN 31
  29
  30/* Create memory pool for log messages */
  31obj_pool_gen(log, char, 4096)
  32
  33static char* log_copy(uint32_t length, char *log)
  34{
  35        char *buffer;
  36        log_free(log_pool.size);
  37        buffer = log_pointer(log_alloc(length));
  38        strncpy(buffer, log, length);
  39        return buffer;
  40}
  41
  42static struct {
  43        uint32_t action, propLength, textLength, srcRev, mark, type;
  44        uint32_t src[REPO_MAX_PATH_DEPTH], dst[REPO_MAX_PATH_DEPTH];
  45        uint32_t text_delta, prop_delta;
  46} node_ctx;
  47
  48static struct {
  49        uint32_t revision, author;
  50        unsigned long timestamp;
  51        char *log;
  52} rev_ctx;
  53
  54static struct {
  55        uint32_t version, uuid, url;
  56} dump_ctx;
  57
  58static struct {
  59        uint32_t svn_log, svn_author, svn_date, svn_executable, svn_special, uuid,
  60                revision_number, node_path, node_kind, node_action,
  61                node_copyfrom_path, node_copyfrom_rev, text_content_length,
  62                prop_content_length, content_length, svn_fs_dump_format_version,
  63                /* version 3 format */
  64                text_delta, prop_delta;
  65} keys;
  66
  67static void reset_node_ctx(char *fname)
  68{
  69        node_ctx.type = 0;
  70        node_ctx.action = NODEACT_UNKNOWN;
  71        node_ctx.propLength = LENGTH_UNKNOWN;
  72        node_ctx.textLength = LENGTH_UNKNOWN;
  73        node_ctx.src[0] = ~0;
  74        node_ctx.srcRev = 0;
  75        pool_tok_seq(REPO_MAX_PATH_DEPTH, node_ctx.dst, "/", fname);
  76        node_ctx.mark = 0;
  77        node_ctx.text_delta = 0;
  78        node_ctx.prop_delta = 0;
  79}
  80
  81static void reset_rev_ctx(uint32_t revision)
  82{
  83        rev_ctx.revision = revision;
  84        rev_ctx.timestamp = 0;
  85        rev_ctx.log = NULL;
  86        rev_ctx.author = ~0;
  87}
  88
  89static void reset_dump_ctx(uint32_t url)
  90{
  91        dump_ctx.url = url;
  92        dump_ctx.version = 1;
  93        dump_ctx.uuid = ~0;
  94}
  95
  96static void init_keys(void)
  97{
  98        keys.svn_log = pool_intern("svn:log");
  99        keys.svn_author = pool_intern("svn:author");
 100        keys.svn_date = pool_intern("svn:date");
 101        keys.svn_executable = pool_intern("svn:executable");
 102        keys.svn_special = pool_intern("svn:special");
 103        keys.uuid = pool_intern("UUID");
 104        keys.revision_number = pool_intern("Revision-number");
 105        keys.node_path = pool_intern("Node-path");
 106        keys.node_kind = pool_intern("Node-kind");
 107        keys.node_action = pool_intern("Node-action");
 108        keys.node_copyfrom_path = pool_intern("Node-copyfrom-path");
 109        keys.node_copyfrom_rev = pool_intern("Node-copyfrom-rev");
 110        keys.text_content_length = pool_intern("Text-content-length");
 111        keys.prop_content_length = pool_intern("Prop-content-length");
 112        keys.content_length = pool_intern("Content-length");
 113        keys.svn_fs_dump_format_version = pool_intern("SVN-fs-dump-format-version");
 114        /* version 3 format (Subversion 1.1.0) */
 115        keys.text_delta = pool_intern("Text-delta");
 116        keys.prop_delta = pool_intern("Prop-delta");
 117}
 118
 119static void read_props(void)
 120{
 121        uint32_t len;
 122        uint32_t key = ~0;
 123        char *val = NULL;
 124        char *t;
 125        while ((t = buffer_read_line()) && strcmp(t, "PROPS-END")) {
 126                if (!strncmp(t, "K ", 2)) {
 127                        len = atoi(&t[2]);
 128                        key = pool_intern(buffer_read_string(len));
 129                        buffer_read_line();
 130                } else if (!strncmp(t, "V ", 2)) {
 131                        len = atoi(&t[2]);
 132                        val = buffer_read_string(len);
 133                        if (key == keys.svn_log) {
 134                                /* Value length excludes terminating nul. */
 135                                rev_ctx.log = log_copy(len + 1, val);
 136                        } else if (key == keys.svn_author) {
 137                                rev_ctx.author = pool_intern(val);
 138                        } else if (key == keys.svn_date) {
 139                                if (parse_date_basic(val, &rev_ctx.timestamp, NULL))
 140                                        fprintf(stderr, "Invalid timestamp: %s\n", val);
 141                        } else if (key == keys.svn_executable) {
 142                                node_ctx.type = REPO_MODE_EXE;
 143                        } else if (key == keys.svn_special) {
 144                                node_ctx.type = REPO_MODE_LNK;
 145                        }
 146                        key = ~0;
 147                        buffer_read_line();
 148                }
 149        }
 150}
 151
 152static void handle_node(void)
 153{
 154        uint32_t old_mode = 0;
 155
 156        if (node_ctx.text_delta || node_ctx.prop_delta)
 157                die("text and property deltas not supported");
 158
 159        if (node_ctx.propLength != LENGTH_UNKNOWN && node_ctx.propLength)
 160                read_props();
 161
 162        if (node_ctx.srcRev)
 163                old_mode = repo_copy(node_ctx.srcRev, node_ctx.src, node_ctx.dst);
 164
 165        if (node_ctx.textLength != LENGTH_UNKNOWN &&
 166            node_ctx.type != REPO_MODE_DIR)
 167                node_ctx.mark = next_blob_mark();
 168
 169        if (node_ctx.action == NODEACT_DELETE) {
 170                repo_delete(node_ctx.dst);
 171        } else if (node_ctx.action == NODEACT_CHANGE ||
 172                           node_ctx.action == NODEACT_REPLACE) {
 173                if (node_ctx.action == NODEACT_REPLACE &&
 174                    node_ctx.type == REPO_MODE_DIR)
 175                        repo_replace(node_ctx.dst, node_ctx.mark);
 176                else if (node_ctx.propLength != LENGTH_UNKNOWN)
 177                        repo_modify(node_ctx.dst, node_ctx.type, node_ctx.mark);
 178                else if (node_ctx.textLength != LENGTH_UNKNOWN)
 179                        old_mode = repo_replace(node_ctx.dst, node_ctx.mark);
 180        } else if (node_ctx.action == NODEACT_ADD) {
 181                if (node_ctx.srcRev && node_ctx.propLength != LENGTH_UNKNOWN)
 182                        repo_modify(node_ctx.dst, node_ctx.type, node_ctx.mark);
 183                else if (node_ctx.srcRev && node_ctx.textLength != LENGTH_UNKNOWN)
 184                        old_mode = repo_replace(node_ctx.dst, node_ctx.mark);
 185                else if ((node_ctx.type == REPO_MODE_DIR && !node_ctx.srcRev) ||
 186                         node_ctx.textLength != LENGTH_UNKNOWN)
 187                        repo_add(node_ctx.dst, node_ctx.type, node_ctx.mark);
 188        }
 189
 190        if (node_ctx.propLength == LENGTH_UNKNOWN && old_mode)
 191                node_ctx.type = old_mode;
 192
 193        if (node_ctx.mark)
 194                fast_export_blob(node_ctx.type, node_ctx.mark, node_ctx.textLength);
 195        else if (node_ctx.textLength != LENGTH_UNKNOWN)
 196                buffer_skip_bytes(node_ctx.textLength);
 197}
 198
 199static void handle_revision(void)
 200{
 201        if (rev_ctx.revision)
 202                repo_commit(rev_ctx.revision, rev_ctx.author, rev_ctx.log,
 203                        dump_ctx.uuid, dump_ctx.url, rev_ctx.timestamp);
 204}
 205
 206void svndump_read(const char *url)
 207{
 208        char *val;
 209        char *t;
 210        uint32_t active_ctx = DUMP_CTX;
 211        uint32_t len;
 212        uint32_t key;
 213
 214        reset_dump_ctx(pool_intern(url));
 215        while ((t = buffer_read_line())) {
 216                val = strstr(t, ": ");
 217                if (!val)
 218                        continue;
 219                *val++ = '\0';
 220                *val++ = '\0';
 221                key = pool_intern(t);
 222
 223                if (key == keys.svn_fs_dump_format_version) {
 224                        dump_ctx.version = atoi(val);
 225                        if (dump_ctx.version > 3)
 226                                die("expected svn dump format version <= 3, found %d",
 227                                    dump_ctx.version);
 228                } else if (key == keys.uuid) {
 229                        dump_ctx.uuid = pool_intern(val);
 230                } else if (key == keys.revision_number) {
 231                        if (active_ctx == NODE_CTX)
 232                                handle_node();
 233                        if (active_ctx != DUMP_CTX)
 234                                handle_revision();
 235                        active_ctx = REV_CTX;
 236                        reset_rev_ctx(atoi(val));
 237                } else if (key == keys.node_path) {
 238                        if (active_ctx == NODE_CTX)
 239                                handle_node();
 240                        active_ctx = NODE_CTX;
 241                        reset_node_ctx(val);
 242                } else if (key == keys.node_kind) {
 243                        if (!strcmp(val, "dir"))
 244                                node_ctx.type = REPO_MODE_DIR;
 245                        else if (!strcmp(val, "file"))
 246                                node_ctx.type = REPO_MODE_BLB;
 247                        else
 248                                fprintf(stderr, "Unknown node-kind: %s\n", val);
 249                } else if (key == keys.node_action) {
 250                        if (!strcmp(val, "delete")) {
 251                                node_ctx.action = NODEACT_DELETE;
 252                        } else if (!strcmp(val, "add")) {
 253                                node_ctx.action = NODEACT_ADD;
 254                        } else if (!strcmp(val, "change")) {
 255                                node_ctx.action = NODEACT_CHANGE;
 256                        } else if (!strcmp(val, "replace")) {
 257                                node_ctx.action = NODEACT_REPLACE;
 258                        } else {
 259                                fprintf(stderr, "Unknown node-action: %s\n", val);
 260                                node_ctx.action = NODEACT_UNKNOWN;
 261                        }
 262                } else if (key == keys.node_copyfrom_path) {
 263                        pool_tok_seq(REPO_MAX_PATH_DEPTH, node_ctx.src, "/", val);
 264                } else if (key == keys.node_copyfrom_rev) {
 265                        node_ctx.srcRev = atoi(val);
 266                } else if (key == keys.text_content_length) {
 267                        node_ctx.textLength = atoi(val);
 268                } else if (key == keys.prop_content_length) {
 269                        node_ctx.propLength = atoi(val);
 270                } else if (key == keys.text_delta) {
 271                        node_ctx.text_delta = !strcmp(val, "true");
 272                } else if (key == keys.prop_delta) {
 273                        node_ctx.prop_delta = !strcmp(val, "true");
 274                } else if (key == keys.content_length) {
 275                        len = atoi(val);
 276                        buffer_read_line();
 277                        if (active_ctx == REV_CTX) {
 278                                read_props();
 279                        } else if (active_ctx == NODE_CTX) {
 280                                handle_node();
 281                                active_ctx = REV_CTX;
 282                        } else {
 283                                fprintf(stderr, "Unexpected content length header: %"PRIu32"\n", len);
 284                                buffer_skip_bytes(len);
 285                        }
 286                }
 287        }
 288        if (active_ctx == NODE_CTX)
 289                handle_node();
 290        if (active_ctx != DUMP_CTX)
 291                handle_revision();
 292}
 293
 294int svndump_init(const char *filename)
 295{
 296        if (buffer_init(filename))
 297                return error("cannot open %s: %s", filename, strerror(errno));
 298        repo_init();
 299        reset_dump_ctx(~0);
 300        reset_rev_ctx(0);
 301        reset_node_ctx(NULL);
 302        init_keys();
 303        return 0;
 304}
 305
 306void svndump_deinit(void)
 307{
 308        log_reset();
 309        repo_reset();
 310        reset_dump_ctx(~0);
 311        reset_rev_ctx(0);
 312        reset_node_ctx(NULL);
 313        if (buffer_deinit())
 314                fprintf(stderr, "Input error\n");
 315        if (ferror(stdout))
 316                fprintf(stderr, "Output error\n");
 317}
 318
 319void svndump_reset(void)
 320{
 321        log_reset();
 322        buffer_reset();
 323        repo_reset();
 324        reset_dump_ctx(~0);
 325        reset_rev_ctx(0);
 326        reset_node_ctx(NULL);
 327}