diff --git a/swh/loader/svn/loader.py b/swh/loader/svn/loader.py index dede674..9c323f2 100644 --- a/swh/loader/svn/loader.py +++ b/swh/loader/svn/loader.py @@ -1,814 +1,806 @@ # Copyright (C) 2015-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information """Loader in charge of injecting either new or existing svn mirrors to swh-storage. """ from datetime import datetime from mmap import ACCESS_WRITE, mmap import os import pty import re import shutil from subprocess import Popen import tempfile -from typing import Dict, Iterator, List, Optional, Tuple +from typing import Dict, Iterator, List, Optional, Sequence, Tuple from subvertpy import SubversionException from swh.loader.core.loader import BaseLoader from swh.loader.core.utils import clean_dangling_folders from swh.loader.exception import NotFound from swh.loader.svn.svn import SvnRepo from swh.model import from_disk, hashutil from swh.model.model import ( Content, Directory, Origin, Revision, SkippedContent, Snapshot, SnapshotBranch, TargetType, ) from swh.storage.algos.snapshot import snapshot_get_latest from swh.storage.interface import StorageInterface from . import converters from .exception import SvnLoaderHistoryAltered, SvnLoaderUneventful from .utils import ( OutputStream, init_svn_repo_from_archive_dump, init_svn_repo_from_dump, ) DEFAULT_BRANCH = b"HEAD" TEMPORARY_DIR_PREFIX_PATTERN = "swh.loader.svn." SUBVERSION_ERROR = re.compile(r".*(E[0-9]{6}):.*") SUBVERSION_NOT_FOUND = "E170013" class SvnLoader(BaseLoader): - """Swh svn loader. - - The repository is either remote or local. The loader deals with + """SVN loader. The repository is either remote or local. The loader deals with update on an already previously loaded repository. """ visit_type = "svn" def __init__( self, storage: StorageInterface, url: str, origin_url: Optional[str] = None, visit_date: Optional[datetime] = None, - destination_path: Optional[str] = None, - swh_revision: Optional[str] = None, incremental: bool = True, temp_directory: str = "/tmp", debug: bool = False, check_revision: int = 0, max_content_size: Optional[int] = None, ): - """Load an svn repository. + """Load a svn repository (either remote or local). Args: - ... + url: The default origin url + origin_url: Optional original url override to use as origin reference in the + archive. If not provided, "url" is used as origin. + visit_date: Optional date to override the visit date incremental: If True, the default, starts from the last snapshot (if any). Otherwise, starts from the initial commit of the repository. + temp_directory: The temporary directory to use as root directory for working + directory computations + debug: If true, run the loader in debug mode. At the end of the loading, the + temporary working directory is not cleaned up to ease inspection. + Defaults to false. + check_revision: The number of svn commits between checks for hash divergence + max_content_size: Default max content size allowed """ super().__init__( storage=storage, logging_class="swh.loader.svn.SvnLoader", max_content_size=max_content_size, ) # technical svn uri to act on svn repository self.svn_url = url # origin url as unique identifier for origin in swh archive self.origin_url = origin_url if origin_url else self.svn_url self.debug = debug self.temp_directory = temp_directory self.done = False self.svnrepo = None # Revision check is configurable self.check_revision = check_revision # internal state used to store swh objects self._contents: List[Content] = [] self._skipped_contents: List[SkippedContent] = [] self._directories: List[Directory] = [] self._revisions: List[Revision] = [] self._snapshot: Optional[Snapshot] = None # internal state, current visit self._last_revision = None self._visit_status = "full" self._load_status = "uneventful" self.visit_date = visit_date - self.destination_path = destination_path self.incremental = incremental self.snapshot: Optional[Snapshot] = None # state from previous visit self.latest_snapshot = None self.latest_revision = None def pre_cleanup(self): """Cleanup potential dangling files from prior runs (e.g. OOM killed tasks) """ clean_dangling_folders( self.temp_directory, pattern_check=TEMPORARY_DIR_PREFIX_PATTERN, log=self.log, ) def cleanup(self): """Clean up the svn repository's working representation on disk. """ if not self.svnrepo: # could happen if `prepare` fails return if self.debug: self.log.error( """NOT FOR PRODUCTION - debug flag activated Local repository not cleaned up for investigation: %s""", self.svnrepo.local_url.decode("utf-8"), ) return self.svnrepo.clean_fs() - def swh_revision_hash_tree_at_svn_revision(self, revision): + def swh_revision_hash_tree_at_svn_revision(self, revision: int) -> bytes: """Compute and return the hash tree at a given svn revision. Args: - rev (int): the svn revision we want to check + rev: the svn revision we want to check Returns: The hash tree directory as bytes. """ + assert self.svnrepo is not None local_dirname, local_url = self.svnrepo.export_temporary(revision) h = from_disk.Directory.from_disk(path=local_url).hash self.svnrepo.clean_fs(local_dirname) return h def _latest_snapshot_revision( self, origin_url: str, ) -> Optional[Tuple[Snapshot, Revision]]: """Look for latest snapshot revision and returns it if any. Args: origin_url: Origin identifier previous_swh_revision: possible previous swh revision (either a dict or revision identifier) Returns: Tuple of the latest Snapshot from the previous visit and its targeted revision if any or None otherwise. """ storage = self.storage latest_snapshot = snapshot_get_latest(storage, origin_url) if not latest_snapshot: return None branches = latest_snapshot.branches if not branches: return None branch = branches.get(DEFAULT_BRANCH) if not branch: return None if branch.target_type != TargetType.REVISION: return None swh_id = branch.target revision = storage.revision_get([swh_id])[0] if not revision: return None return latest_snapshot, revision - def build_swh_revision(self, rev, commit, dir_id, parents): + def build_swh_revision( + self, rev: int, commit: Dict, dir_id: bytes, parents: Sequence[bytes] + ) -> Revision: """Build the swh revision dictionary. This adds: - the `'synthetic`' flag to true - the '`extra_headers`' containing the repository's uuid and the svn revision number. Args: - rev (int): the svn revision number - commit (dict): the commit data: revision id, date, author, and message - dir_id (bytes): the upper tree's hash identifier - parents ([bytes]): the parents' identifiers + rev: the svn revision number + commit: the commit data: revision id, date, author, and message + dir_id: the upper tree's hash identifier + parents: the parents' identifiers Returns: The swh revision corresponding to the svn revision. """ + assert self.svnrepo is not None return converters.build_swh_revision( rev, commit, self.svnrepo.uuid, dir_id, parents ) def check_history_not_altered( - self, svnrepo, revision_start: int, swh_rev: Revision + self, svnrepo: SvnRepo, revision_start: int, swh_rev: Revision ) -> bool: """Given a svn repository, check if the history was modified in between visits. """ revision_id = swh_rev.id parents = swh_rev.parents hash_data_per_revs = svnrepo.swh_hash_data_at_revision(revision_start) rev = revision_start rev, _, commit, _, root_dir = list(hash_data_per_revs)[0] dir_id = root_dir.hash swh_revision = self.build_swh_revision(rev, commit, dir_id, parents) swh_revision_id = swh_revision.id return swh_revision_id == revision_id def start_from(self) -> Tuple[int, int, Dict[int, Tuple[bytes, ...]]]: """Determine from where to start the loading. Returns: tuple (revision_start, revision_end, revision_parents) Raises: SvnLoaderHistoryAltered: When a hash divergence has been detected (should not happen) SvnLoaderUneventful: Nothing changed since last visit """ assert self.svnrepo is not None, "svnrepo initialized in the `prepare` method" revision_head = self.svnrepo.head_revision() if revision_head == 0: # empty repository case revision_start = 0 revision_end = 0 else: # default configuration revision_start = self.svnrepo.initial_revision() revision_end = revision_head revision_parents: Dict[int, Tuple[bytes, ...]] = {revision_start: ()} # start from a previous revision if any if self.incremental and self.latest_revision is not None: extra_headers = dict(self.latest_revision.extra_headers) revision_start = int(extra_headers[b"svn_revision"]) revision_parents = { revision_start: self.latest_revision.parents, } self.log.debug( "svn export --ignore-keywords %s@%s", self.svnrepo.remote_url, revision_start, ) if not self.check_history_not_altered( self.svnrepo, revision_start, self.latest_revision ): self.log.debug( ( "History of svn %s@%s altered. " "A complete reloading of the repository will be performed." ), self.svnrepo.remote_url, revision_start, ) revision_start = self.svnrepo.initial_revision() # now we know history is ok, we start at next revision revision_start = revision_start + 1 # and the parent become the latest know revision for # that repository revision_parents[revision_start] = (self.latest_revision.id,) if revision_start > revision_end: msg = "%s@%s already injected." % (self.svnrepo.remote_url, revision_end) raise SvnLoaderUneventful(msg) self.log.info( "Processing revisions [%s-%s] for %s", revision_start, revision_end, self.svnrepo, ) return revision_start, revision_end, revision_parents - def _check_revision_divergence(self, count, rev, dir_id): + def _check_revision_divergence(self, rev: int, dir_id: bytes) -> None: """Check for hash revision computation divergence. - The Rationale behind this is that svn can trigger unknown - edge cases (mixed CRLF, svn properties, etc...). Those are - not always easy to spot. Adding a check will help in - spotting missing edge cases. + The Rationale behind this is that svn can trigger unknown edge cases (mixed + CRLF, svn properties, etc...). Those are not always easy to spot. Adding a + regular check will help spotting potential missing edge cases. Args: - count (int): The number of revisions done so far - rev (dict): The actual revision we are computing from - dir_id (bytes): The actual directory for the given revision - - Returns: - False if no hash divergence detected + rev: The actual revision we are computing from + dir_id: The actual directory for the given revision Raises ValueError if a hash divergence is detected """ - # hash computation check - if (self.check_revision != 0 and count % self.check_revision) == 0: - self.log.debug("Checking hash computations on revision %s...", rev) - checked_dir_id = self.swh_revision_hash_tree_at_svn_revision(rev) - if checked_dir_id != dir_id: - err = ( - "Hash tree computation divergence detected " - "(%s != %s), stopping!" - % ( - hashutil.hash_to_hex(dir_id), - hashutil.hash_to_hex(checked_dir_id), - ) - ) - raise ValueError(err) + + self.log.debug("Checking hash computations on revision %s...", rev) + checked_dir_id = self.swh_revision_hash_tree_at_svn_revision(rev) + if checked_dir_id != dir_id: + err = ( + "Hash tree computation divergence detected " + "(%s != %s), stopping!" + % (hashutil.hash_to_hex(dir_id), hashutil.hash_to_hex(checked_dir_id),) + ) + raise ValueError(err) def process_svn_revisions( self, svnrepo, revision_start, revision_end, revision_parents ) -> Iterator[ Tuple[List[Content], List[SkippedContent], List[Directory], Revision] ]: """Process svn revisions from revision_start to revision_end. At each svn revision, apply new diffs and simultaneously compute swh hashes. This yields those computed swh hashes as a tuple (contents, directories, revision). Note that at every `self.check_revision`, a supplementary check takes place to check for hash-tree divergence (related T570). Yields: tuple (contents, directories, revision) of dict as a dictionary with keys, sha1_git, sha1, etc... Raises: ValueError in case of a hash divergence detection """ gen_revs = svnrepo.swh_hash_data_per_revision(revision_start, revision_end) swh_revision = None count = 0 for rev, nextrev, commit, new_objects, root_directory in gen_revs: count += 1 # Send the associated contents/directories _contents, _skipped_contents, _directories = new_objects # compute the fs tree's checksums dir_id = root_directory.hash swh_revision = self.build_swh_revision( rev, commit, dir_id, revision_parents.get(rev, ()) ) self.log.debug( "rev: %s, swhrev: %s, dir: %s", rev, hashutil.hash_to_hex(swh_revision.id), hashutil.hash_to_hex(dir_id), ) - if self.check_revision: - self._check_revision_divergence(count, rev, dir_id) + if ( + self.check_revision + and self.check_revision != 0 + and count % self.check_revision == 0 + ): + self._check_revision_divergence(rev, dir_id) if nextrev: revision_parents[nextrev] = [swh_revision.id] yield _contents, _skipped_contents, _directories, swh_revision def prepare_origin_visit(self): self.origin = Origin(url=self.origin_url if self.origin_url else self.svn_url) def prepare(self): latest_snapshot_revision = self._latest_snapshot_revision(self.origin_url) if latest_snapshot_revision: self.latest_snapshot, self.latest_revision = latest_snapshot_revision - if self.destination_path: - local_dirname = self.destination_path - else: - local_dirname = tempfile.mkdtemp( - suffix="-%s" % os.getpid(), - prefix=TEMPORARY_DIR_PREFIX_PATTERN, - dir=self.temp_directory, - ) + local_dirname = tempfile.mkdtemp( + suffix="-%s" % os.getpid(), + prefix=TEMPORARY_DIR_PREFIX_PATTERN, + dir=self.temp_directory, + ) try: self.svnrepo = SvnRepo( self.svn_url, self.origin_url, local_dirname, self.max_content_size ) except SubversionException as e: error_msgs = [ "Unable to connect to a repository at URL", "Unknown URL type", ] for msg in error_msgs: if msg in e.args[0]: self._load_status = "uneventful" raise NotFound(e) raise try: revision_start, revision_end, revision_parents = self.start_from() self.swh_revision_gen = self.process_svn_revisions( self.svnrepo, revision_start, revision_end, revision_parents ) except SvnLoaderUneventful as e: self.log.warning(e) if self.latest_snapshot: self._snapshot = self.latest_snapshot self.done = True self._load_status = "uneventful" except SvnLoaderHistoryAltered as e: self.log.error(e) self.done = True self._visit_status = "partial" def fetch_data(self): """Fetching svn revision information. This will apply svn revision as patch on disk, and at the same time, compute the swh hashes. In effect, fetch_data fetches those data and compute the necessary swh objects. It's then stored in the internal state instance variables (initialized in `_prepare_state`). This is up to `store_data` to actually discuss with the storage to store those objects. Returns: bool: True to continue fetching data (next svn revision), False to stop. """ data = None if self.done: return False try: data = next(self.swh_revision_gen) self._load_status = "eventful" except StopIteration: self.done = True self._visit_status = "full" return False # Stopping iteration except Exception as e: # svn:external, hash divergence, i/o error... self.log.exception(e) self.done = True self._visit_status = "partial" return False # Stopping iteration self._contents, self._skipped_contents, self._directories, rev = data if rev: self._last_revision = rev self._revisions.append(rev) return True # next svn revision def store_data(self): """We store the data accumulated in internal instance variable. If the iteration over the svn revisions is done, we create the snapshot and flush to storage the data. This also resets the internal instance variable state. """ self.storage.skipped_content_add(self._skipped_contents) self.storage.content_add(self._contents) self.storage.directory_add(self._directories) self.storage.revision_add(self._revisions) if self.done: # finish line, snapshot! self.snapshot = self.generate_and_load_snapshot( revision=self._last_revision, snapshot=self._snapshot ) self.flush() self.loaded_snapshot_id = self.snapshot.id # reset internal state for next iteration self._revisions = [] def generate_and_load_snapshot( self, revision: Optional[Revision] = None, snapshot: Optional[Snapshot] = None ) -> Snapshot: """Create the snapshot either from existing revision or snapshot. Revision (supposedly new) has priority over the snapshot (supposedly existing one). Args: revision (dict): Last revision seen if any (None by default) snapshot (dict): Snapshot to use if any (None by default) Returns: Optional[Snapshot] The newly created snapshot """ if revision: # Priority to the revision snap = Snapshot( branches={ DEFAULT_BRANCH: SnapshotBranch( target=revision.id, target_type=TargetType.REVISION ) } ) elif snapshot: # Fallback to prior snapshot snap = snapshot else: raise ValueError( "generate_and_load_snapshot called with null revision and snapshot!" ) self.log.debug("snapshot: %s", snap) self.storage.snapshot_add([snap]) return snap def load_status(self): return { "status": self._load_status, } def visit_status(self): return self._visit_status def post_load(self, success: bool = True) -> None: if success and self._last_revision is not None: - # force revision divergence check - self.check_revision = 1 # check if the reconstructed filesystem for the last loaded revision is - # consistent with the one obtained with a svn export operation, if it is - # not an exception will be raised to report the issue and mark the visit - # as partial + # consistent with the one obtained with a svn export operation. If it is not + # the case, an exception will be raised to report the issue and mark the + # visit as partial self._check_revision_divergence( - self.check_revision, int(dict(self._last_revision.extra_headers)[b"svn_revision"]), self._last_revision.directory, ) class SvnLoaderFromDumpArchive(SvnLoader): - """Uncompress an archive containing an svn dump, mount the svn dump as - an svn repository and load said repository. + """Uncompress an archive containing an svn dump, mount the svn dump as a local svn + repository and load that repository. """ def __init__( self, storage: StorageInterface, url: str, archive_path: str, origin_url: Optional[str] = None, - destination_path: Optional[str] = None, - swh_revision: Optional[str] = None, incremental: bool = False, visit_date: Optional[datetime] = None, temp_directory: str = "/tmp", debug: bool = False, check_revision: int = 0, max_content_size: Optional[int] = None, ): super().__init__( storage=storage, url=url, origin_url=origin_url, - destination_path=destination_path, - swh_revision=swh_revision, incremental=incremental, visit_date=visit_date, temp_directory=temp_directory, debug=debug, check_revision=check_revision, max_content_size=max_content_size, ) self.archive_path = archive_path self.temp_dir = None self.repo_path = None def prepare(self): self.log.info("Archive to mount and load %s", self.archive_path) self.temp_dir, self.repo_path = init_svn_repo_from_archive_dump( self.archive_path, prefix=TEMPORARY_DIR_PREFIX_PATTERN, suffix="-%s" % os.getpid(), root_dir=self.temp_directory, ) self.svn_url = f"file://{self.repo_path}" super().prepare() def cleanup(self): super().cleanup() if self.temp_dir and os.path.exists(self.temp_dir): self.log.debug( "Clean up temporary directory dump %s for project %s", self.temp_dir, os.path.basename(self.repo_path), ) shutil.rmtree(self.temp_dir) class SvnLoaderFromRemoteDump(SvnLoader): - """ - Create a subversion repository dump using the svnrdump utility, - mount it locally and load the repository from it. + """Create a subversion repository dump out of a remote svn repository (using the + svnrdump utility). Then, mount the repository locally and load that repository. + """ def __init__( self, storage: StorageInterface, url: str, origin_url: Optional[str] = None, - destination_path: Optional[str] = None, - swh_revision: Optional[str] = None, incremental: bool = True, visit_date: Optional[datetime] = None, temp_directory: str = "/tmp", debug: bool = False, check_revision: int = 0, max_content_size: Optional[int] = None, ): super().__init__( storage=storage, url=url, origin_url=origin_url, - destination_path=destination_path, - swh_revision=swh_revision, incremental=incremental, visit_date=visit_date, temp_directory=temp_directory, debug=debug, check_revision=check_revision, max_content_size=max_content_size, ) self.temp_dir = tempfile.mkdtemp(dir=self.temp_directory) self.repo_path = None self.truncated_dump = False def get_last_loaded_svn_rev(self, svn_url: str) -> int: """Check if the svn repository has already been visited and return the last loaded svn revision number or -1 otherwise. """ origin = list(self.storage.origin_get([svn_url]))[0] if not origin: return -1 svn_revision = -1 try: latest_snapshot_revision = self._latest_snapshot_revision(origin.url) if latest_snapshot_revision: _, latest_revision = latest_snapshot_revision latest_revision_headers = dict(latest_revision.extra_headers) svn_revision = int(latest_revision_headers[b"svn_revision"]) except Exception: pass return svn_revision - def dump_svn_revisions(self, svn_url, last_loaded_svn_rev=-1): - """ - Generate a subversion dump file using the svnrdump tool. - If the svnrdump command failed somehow, - the produced dump file is analyzed to determine if a partial - loading is still feasible. + def dump_svn_revisions(self, svn_url: str, last_loaded_svn_rev: int = -1) -> str: + """Generate a subversion dump file using the svnrdump tool. If the svnrdump + command failed somehow, the produced dump file is analyzed to determine if a + partial loading is still feasible. Raises: NotFound when the repository is no longer found at url + + Returns: + The dump_path of the repository mounted + """ # Build the svnrdump command line svnrdump_cmd = ["svnrdump", "dump", svn_url] # Launch the svnrdump command while capturing stderr as # successfully dumped revision numbers are printed to it dump_temp_dir = tempfile.mkdtemp(dir=self.temp_dir) dump_name = "".join(c for c in svn_url if c.isalnum()) dump_path = "%s/%s.svndump" % (dump_temp_dir, dump_name) stderr_lines = [] self.log.debug("Executing %s", " ".join(svnrdump_cmd)) with open(dump_path, "wb") as dump_file: stderr_r, stderr_w = pty.openpty() svnrdump = Popen(svnrdump_cmd, stdout=dump_file, stderr=stderr_w) os.close(stderr_w) stderr_stream = OutputStream(stderr_r) readable = True error_codes: List[str] = [] error_messages: List[str] = [] while readable: lines, readable = stderr_stream.read_lines() stderr_lines += lines for line in lines: self.log.debug(line) match = SUBVERSION_ERROR.search(line) if match: error_codes.append(match.group(1)) error_messages.append(line) svnrdump.wait() os.close(stderr_r) if svnrdump.returncode == 0: return dump_path # There was an error but it does not mean that no revisions # can be loaded. # Get the stderr line with latest dumped revision last_dumped_rev = None if len(stderr_lines) > 1: last_dumped_rev = stderr_lines[-2] if last_dumped_rev: # Get the latest dumped revision number matched_rev = re.search(".*revision ([0-9]+)", last_dumped_rev) last_dumped_rev = int(matched_rev.group(1)) if matched_rev else -1 # Check if revisions inside the dump file can be loaded anyway if last_dumped_rev > last_loaded_svn_rev: self.log.debug( ( "svnrdump did not dump all expected revisions " "but revisions range %s:%s are available in " "the generated dump file and will be loaded " "into the archive." ), last_loaded_svn_rev + 1, last_dumped_rev, ) # Truncate the dump file after the last successfully dumped # revision to avoid the loading of corrupted data self.log.debug( ( "Truncating dump file after the last " "successfully dumped revision (%s) to avoid " "the loading of corrupted data" ), last_dumped_rev, ) with open(dump_path, "r+b") as f: with mmap(f.fileno(), 0, access=ACCESS_WRITE) as s: pattern = ( "Revision-number: %s" % (last_dumped_rev + 1) ).encode() n = s.rfind(pattern) if n != -1: s.resize(n) self.truncated_dump = True return dump_path elif last_dumped_rev != -1: raise Exception( ( "Last dumped subversion revision (%s) is " "lesser than the last one loaded into the " "archive (%s)." ) % (last_dumped_rev, last_loaded_svn_rev) ) if SUBVERSION_NOT_FOUND in error_codes: raise NotFound( f"{SUBVERSION_NOT_FOUND}: Repository never existed or disappeared" ) raise Exception( "An error occurred when running svnrdump and " "no exploitable dump file has been generated.\n" + "\n".join(error_messages) ) def prepare(self): # First, check if previous revisions have been loaded for the # subversion origin and get the number of the last one last_loaded_svn_rev = self.get_last_loaded_svn_rev(self.svn_url) # Then try to generate a dump file containing relevant svn revisions # to load, an exception will be thrown if something wrong happened dump_path = self.dump_svn_revisions(self.svn_url, last_loaded_svn_rev) # Finally, mount the dump and load the repository self.log.debug('Mounting dump file with "svnadmin load".') _, self.repo_path = init_svn_repo_from_dump( dump_path, prefix=TEMPORARY_DIR_PREFIX_PATTERN, suffix="-%s" % os.getpid(), root_dir=self.temp_dir, ) self.svn_url = "file://%s" % self.repo_path super().prepare() def cleanup(self): super().cleanup() if self.temp_dir and os.path.exists(self.temp_dir): shutil.rmtree(self.temp_dir) def visit_status(self): if self.truncated_dump: return "partial" else: return super().visit_status() diff --git a/swh/loader/svn/svn.py b/swh/loader/svn/svn.py index f770ea7..6a29d86 100644 --- a/swh/loader/svn/svn.py +++ b/swh/loader/svn/svn.py @@ -1,271 +1,277 @@ # Copyright (C) 2015-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information """SVN client in charge of iterating over svn logs and yield commit representations including the hash tree/content computations per svn commit. """ import logging import os import shutil import tempfile from subvertpy import client, properties from subvertpy.ra import Auth, RemoteAccess, get_username_provider from swh.model.from_disk import Directory from . import converters, ra # When log message contains empty data DEFAULT_AUTHOR_MESSAGE = "" logger = logging.getLogger(__name__) class SvnRepo: """Svn repository representation. Args: - remote_url (str): - origin_url (str): Associated origin identifier - local_dirname (str): Path to write intermediary svn action results + remote_url: Remove svn repository url + origin_url: Associated origin identifier + local_dirname: Path to write intermediary svn action results """ - def __init__(self, remote_url, origin_url, local_dirname, max_content_length): + def __init__( + self, + remote_url: str, + origin_url: str, + local_dirname: str, + max_content_length: int, + ): self.remote_url = remote_url.rstrip("/") self.origin_url = origin_url auth = Auth([get_username_provider()]) # one connection for log iteration self.conn_log = RemoteAccess(self.remote_url, auth=auth) # another for replay self.conn = RemoteAccess(self.remote_url, auth=auth) # one client for update operation self.client = client.Client(auth=auth) self.local_dirname = local_dirname local_name = os.path.basename(self.remote_url) self.local_url = os.path.join(self.local_dirname, local_name).encode("utf-8") self.uuid = self.conn.get_uuid().encode("utf-8") self.swhreplay = ra.Replay(conn=self.conn, rootpath=self.local_url) self.max_content_length = max_content_length def __str__(self): return str( { "swh-origin": self.origin_url, "remote_url": self.remote_url, "local_url": self.local_url, "uuid": self.uuid, } ) def head_revision(self): """Retrieve current head revision. """ return self.conn.get_latest_revnum() def initial_revision(self): """Retrieve the initial revision from which the remote url appeared. """ return 1 def convert_commit_message(self, msg): """Simply encode the commit message. Args: msg (str): the commit message to convert. Returns: The transformed message as bytes. """ if isinstance(msg, bytes): return msg return msg.encode("utf-8") def convert_commit_date(self, date): """Convert the message commit date into a timestamp in swh format. The precision is kept. Args: date (str): the commit date to convert. Returns: The transformed date. """ return converters.svn_date_to_swh_date(date) def convert_commit_author(self, author): """Convert the commit author into an swh person. Args: author (str): the commit author to convert. Returns: Person: a model object """ return converters.svn_author_to_swh_person(author) def __to_entry(self, log_entry): changed_paths, rev, revprops, has_children = log_entry author_date = self.convert_commit_date( revprops.get(properties.PROP_REVISION_DATE) ) author = self.convert_commit_author( revprops.get(properties.PROP_REVISION_AUTHOR) ) message = self.convert_commit_message( revprops.get(properties.PROP_REVISION_LOG, DEFAULT_AUTHOR_MESSAGE) ) return { "rev": rev, "author_date": author_date, "author_name": author, "message": message, } def logs(self, revision_start, revision_end): """Stream svn logs between revision_start and revision_end by chunks of block_size logs. Yields revision and associated revision information between the revision start and revision_end. Args: revision_start: the svn revision starting bound revision_end: the svn revision ending bound Yields: tuple: tuple of revisions and logs: - revisions: list of revisions in order - logs: Dictionary with key revision number and value the log entry. The log entry is a dictionary with the following keys: - author_date: date of the commit - author_name: name of the author - message: commit message """ for log_entry in self.conn_log.iter_log( paths=None, start=revision_start, end=revision_end, discover_changed_paths=False, ): yield self.__to_entry(log_entry) def export(self, revision): """Export the repository to a given version. """ self.client.export( self.remote_url, to=self.local_url.decode("utf-8"), rev=revision, ignore_keywords=True, ) def export_temporary(self, revision): """Export the repository to a given revision in a temporary location. This is up to the caller of this function to clean up the temporary location when done (cf. self.clean_fs method) Args: revision: Revision to export at Returns: The tuple local_dirname the temporary location root folder, local_url where the repository was exported. """ local_dirname = tempfile.mkdtemp( prefix="check-revision-%s." % revision, dir=self.local_dirname ) local_name = os.path.basename(self.remote_url) local_url = os.path.join(local_dirname, local_name) self.client.export( self.remote_url, to=local_url, rev=revision, ignore_keywords=True ) return local_dirname, os.fsencode(local_url) def swh_hash_data_per_revision(self, start_revision, end_revision): """Compute swh hash data per each revision between start_revision and end_revision. Args: start_revision: starting revision end_revision: ending revision Yields: tuple (rev, nextrev, commit, objects_per_path) - rev: current revision - nextrev: next revision - commit: commit data (author, date, message) for such revision - objects_per_path: dictionary of path, swh hash data with type """ for commit in self.logs(start_revision, end_revision): rev = commit["rev"] objects = self.swhreplay.compute_objects(rev) if rev == end_revision: nextrev = None else: nextrev = rev + 1 yield rev, nextrev, commit, objects, self.swhreplay.directory def swh_hash_data_at_revision(self, revision): """Compute the hash data at revision. Expected to be used for update only. """ # Update the disk at revision self.export(revision) # Compute the current hashes on disk directory = Directory.from_disk( path=os.fsencode(self.local_url), max_content_length=self.max_content_length ) # Update the replay collaborator with the right state self.swhreplay = ra.Replay( conn=self.conn, rootpath=self.local_url, directory=directory ) # Retrieve the commit information for revision commit = list(self.logs(revision, revision))[0] yield revision, revision + 1, commit, {}, directory def clean_fs(self, local_dirname=None): """Clean up the local working copy. Args: local_dirname (str): Path to remove recursively if provided. Otherwise, remove the temporary upper root tree used for svn repository loading. """ dirname = local_dirname if local_dirname else self.local_dirname if os.path.exists(dirname): logger.debug("cleanup %s", dirname) shutil.rmtree(dirname) diff --git a/swh/loader/svn/tasks.py b/swh/loader/svn/tasks.py index b8695d5..0009255 100644 --- a/swh/loader/svn/tasks.py +++ b/swh/loader/svn/tasks.py @@ -1,115 +1,108 @@ # Copyright (C) 2015-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from datetime import datetime from typing import Optional from celery import shared_task import iso8601 from .loader import SvnLoader, SvnLoaderFromDumpArchive, SvnLoaderFromRemoteDump def convert_to_datetime(date: Optional[str]) -> Optional[datetime]: try: return iso8601.parse_date(date) except Exception: return None @shared_task(name=__name__ + ".LoadSvnRepository") def load_svn( *, url: Optional[str] = None, origin_url: Optional[str] = None, - destination_path: Optional[str] = None, - swh_revision: Optional[str] = None, visit_date: Optional[str] = None, incremental: Optional[bool] = True, ): """Import a svn repository Args: url: (mandatory) svn's repository url to ingest data from origin_url: Optional original url override to use as origin reference in the archive. If not provided, "url" is used as origin. - destination_path: (optional) root directory to locally retrieve svn's data - swh_revision: (optional) extra revision hex to start from. See - swh.loader.svn.SvnLoader.process docstring visit_date: Optional date to override the visit date incremental: If True, the default, starts from the last snapshot (if any). Otherwise, starts from the initial commit of the repository. """ loader = SvnLoader.from_configfile( url=url, origin_url=origin_url, - destination_path=destination_path, - swh_revision=swh_revision, visit_date=convert_to_datetime(visit_date), incremental=incremental, ) return loader.load() @shared_task(name=__name__ + ".MountAndLoadSvnRepository") def load_svn_from_archive( *, url: Optional[str] = None, archive_path: Optional[str] = None, visit_date: Optional[str] = None, incremental: Optional[bool] = True, ): """1. Mount an svn dump from archive as a local svn repository 2. Load it through the svn loader 3. Clean up mounted svn repository archive Args: url: origin url archive_path: Path on disk to the archive holdin the svn repository to ingest visit_date: Optional date to override the visit date incremental: If True, the default, starts from the last snapshot (if any). Otherwise, starts from the initial commit of the repository. """ loader = SvnLoaderFromDumpArchive.from_configfile( url=url, archive_path=archive_path, visit_date=convert_to_datetime(visit_date), incremental=incremental, ) return loader.load() @shared_task(name=__name__ + ".DumpMountAndLoadSvnRepository") def load_svn_from_remote_dump( *, url: Optional[str] = None, origin_url: Optional[str] = None, visit_date: Optional[str] = None, incremental: Optional[bool] = True, ): """1. Mount a remote svn dump as a local svn repository. 2. Load it through the svn loader. 3. Clean up mounted svn repository archive. Args: url: (mandatory) svn's repository url to ingest data from origin_url: Optional original url override to use as origin reference in the archive. If not provided, "url" is used as origin. visit_date: Optional date to override the visit date incremental: If True, the default, starts from the last snapshot (if any). Otherwise, starts from the initial commit of the repository. """ loader = SvnLoaderFromRemoteDump.from_configfile( url=url, origin_url=origin_url, visit_date=convert_to_datetime(visit_date), incremental=incremental, ) return loader.load() diff --git a/swh/loader/svn/tests/test_loader.py b/swh/loader/svn/tests/test_loader.py index ef2a935..1e4df82 100644 --- a/swh/loader/svn/tests/test_loader.py +++ b/swh/loader/svn/tests/test_loader.py @@ -1,1283 +1,1265 @@ # Copyright (C) 2016-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from enum import Enum from io import BytesIO import os import shutil import subprocess from typing import Any, Dict, List import pytest from subvertpy import SubversionException, delta, repos from subvertpy.ra import Auth, RemoteAccess, get_username_provider from typing_extensions import TypedDict from swh.loader.svn.loader import ( SvnLoader, SvnLoaderFromDumpArchive, SvnLoaderFromRemoteDump, ) from swh.loader.svn.svn import SvnRepo from swh.loader.svn.utils import init_svn_repo_from_dump from swh.loader.tests import ( assert_last_visit_matches, check_snapshot, get_stats, prepare_repository_from_archive, ) from swh.model.from_disk import DentryPerms from swh.model.hashutil import hash_to_bytes from swh.model.model import Snapshot, SnapshotBranch, TargetType GOURMET_SNAPSHOT = Snapshot( id=hash_to_bytes("889cacc2731e3312abfb2b1a0c18ade82a949e07"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("4876cb10aec6f708f7466dddf547567b65f6c39c"), target_type=TargetType.REVISION, ) }, ) GOURMET_UPDATES_SNAPSHOT = Snapshot( id=hash_to_bytes("11086d15317014e43d2438b7ffc712c44f1b8afe"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("171dc35522bfd17dda4e90a542a0377fb2fc707a"), target_type=TargetType.REVISION, ) }, ) def test_loader_svn_not_found_no_mock(swh_storage, tmp_path): """Given an unknown repository, the loader visit ends up in status not_found""" repo_url = "unknown-repository" - loader = SvnLoader(swh_storage, repo_url, destination_path=tmp_path) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "uneventful"} assert_last_visit_matches( swh_storage, repo_url, status="not_found", type="svn", ) @pytest.mark.parametrize( "exception_msg", ["Unable to connect to a repository at URL", "Unknown URL type",] ) def test_loader_svn_not_found(swh_storage, tmp_path, exception_msg, mocker): """Given unknown repository issues, the loader visit ends up in status not_found""" mock = mocker.patch("swh.loader.svn.loader.SvnRepo") mock.side_effect = SubversionException(exception_msg, 0) unknown_repo_url = "unknown-repository" - loader = SvnLoader(swh_storage, unknown_repo_url, destination_path=tmp_path) + loader = SvnLoader(swh_storage, unknown_repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "uneventful"} assert_last_visit_matches( swh_storage, unknown_repo_url, status="not_found", type="svn", ) @pytest.mark.parametrize( "exception", [ SubversionException("Irrelevant message, considered a failure", 10), SubversionException("Present but fails to read, considered a failure", 20), ValueError("considered a failure"), ], ) def test_loader_svn_failures(swh_storage, tmp_path, exception, mocker): """Given any errors raised, the loader visit ends up in status failed""" mock = mocker.patch("swh.loader.svn.loader.SvnRepo") mock.side_effect = exception existing_repo_url = "existing-repo-url" - loader = SvnLoader(swh_storage, existing_repo_url, destination_path=tmp_path) + loader = SvnLoader(swh_storage, existing_repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "failed"} assert_last_visit_matches( swh_storage, existing_repo_url, status="failed", type="svn", ) def test_loader_svnrdump_not_found(swh_storage, tmp_path, mocker): """Loading from remote dump which does not exist should end up as not_found visit""" unknown_repo_url = "file:///tmp/svn.code.sf.net/p/white-rats-studios/svn" loader = SvnLoaderFromRemoteDump( - swh_storage, unknown_repo_url, destination_path=tmp_path + swh_storage, unknown_repo_url, temp_directory=tmp_path ) assert loader.load() == {"status": "uneventful"} assert_last_visit_matches( swh_storage, unknown_repo_url, status="not_found", type="svn", ) def test_loader_svnrdump_no_such_revision(swh_storage, tmp_path, datadir): """Visit multiple times an origin with the remote loader should not raise. It used to fail the ingestion on the second visit with a "No such revision x, 160006" message. """ archive_ori_dump = os.path.join(datadir, "penguinsdbtools2018.dump.gz") archive_dump_dir = os.path.join(tmp_path, "dump") os.mkdir(archive_dump_dir) archive_dump = os.path.join(archive_dump_dir, "penguinsdbtools2018.dump.gz") # loader now drops the dump as soon as it's mounted so we need to make a copy first shutil.copyfile(archive_ori_dump, archive_dump) loading_path = str(tmp_path / "loading") os.mkdir(loading_path) # Prepare the dump as a local svn repository for test purposes temp_dir, repo_path = init_svn_repo_from_dump( archive_dump, root_dir=tmp_path, gzip=True ) repo_url = f"file://{repo_path}" - loader = SvnLoaderFromRemoteDump( - swh_storage, repo_url, destination_path=loading_path - ) + loader = SvnLoaderFromRemoteDump(swh_storage, repo_url, temp_directory=loading_path) assert loader.load() == {"status": "eventful"} actual_visit = assert_last_visit_matches( swh_storage, repo_url, status="full", type="svn", ) loader2 = SvnLoaderFromRemoteDump( - swh_storage, repo_url, destination_path=loading_path + swh_storage, repo_url, temp_directory=loading_path ) # Visiting a second time the same repository should be uneventful... assert loader2.load() == {"status": "uneventful"} actual_visit2 = assert_last_visit_matches( swh_storage, repo_url, status="full", type="svn", ) assert actual_visit.snapshot is not None # ... with the same snapshot as the first visit assert actual_visit2.snapshot == actual_visit.snapshot def test_loader_svn_new_visit(swh_storage, datadir, tmp_path): """Eventful visit should yield 1 snapshot""" archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) - loader = SvnLoader(swh_storage, repo_url, destination_path=tmp_path) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) stats = get_stats(loader.storage) assert stats == { "content": 19, "directory": 17, "origin": 1, "origin_visit": 1, "release": 0, "revision": 6, "skipped_content": 0, "snapshot": 1, } check_snapshot(GOURMET_SNAPSHOT, loader.storage) def test_loader_svn_2_visits_no_change(swh_storage, datadir, tmp_path): """Visit multiple times a repository with no change should yield the same snapshot """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} visit_status1 = assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) assert loader.load() == {"status": "uneventful"} visit_status2 = assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) assert visit_status1.date < visit_status2.date assert visit_status1.snapshot == visit_status2.snapshot stats = get_stats(loader.storage) assert stats["origin_visit"] == 1 + 1 # computed twice the same snapshot assert stats["snapshot"] == 1 # even starting from previous revision... start_revision = loader.storage.revision_get( [hash_to_bytes("95edacc8848369d6fb1608e887d6d2474fd5224f")] )[0] assert start_revision is not None - loader = SvnLoader(swh_storage, repo_url, swh_revision=start_revision) + loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "uneventful"} stats = get_stats(loader.storage) assert stats["origin_visit"] == 2 + 1 # ... with no change in repository, this yields the same snapshot assert stats["snapshot"] == 1 assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) def test_loader_tampered_repository(swh_storage, datadir, tmp_path): """In this scenario, the dump has been tampered with to modify the commit log [1]. This results in a hash divergence which is detected at startup after a new run for the same origin. In effect, this will perform a complete reloading of the repository. [1] Tampering with revision 6 log message following: ``` tar xvf pkg-gourmet.tgz # initial repository ingested cd pkg-gourmet/ echo "Tampering with commit log message for fun and profit" > log.txt svnadmin setlog . -r 6 log.txt --bypass-hooks tar cvf pkg-gourmet-tampered-rev6-log.tgz pkg-gourmet/ ``` """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} check_snapshot(GOURMET_SNAPSHOT, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) archive_path2 = os.path.join(datadir, "pkg-gourmet-tampered-rev6-log.tgz") repo_tampered_url = prepare_repository_from_archive( archive_path2, archive_name, tmp_path ) loader2 = SvnLoader(swh_storage, repo_tampered_url, origin_url=repo_url) assert loader2.load() == {"status": "eventful"} assert_last_visit_matches( loader2.storage, repo_url, status="full", type="svn", snapshot=hash_to_bytes("c499eebc1e201024d47d24053ac0080049305897"), ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 2 assert stats["snapshot"] == 2 def test_loader_svn_visit_with_changes(swh_storage, datadir, tmp_path): """In this scenario, the repository has been updated with new changes. The loading visit should result in new objects stored and 1 new snapshot. """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_initial_url = prepare_repository_from_archive( archive_path, archive_name, tmp_path ) # repo_initial_url becomes the origin_url we want to visit some more below loader = SvnLoader(swh_storage, repo_initial_url) assert loader.load() == {"status": "eventful"} visit_status1 = assert_last_visit_matches( loader.storage, repo_initial_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) archive_path = os.path.join(datadir, "pkg-gourmet-with-updates.tgz") repo_updated_url = prepare_repository_from_archive( archive_path, "pkg-gourmet", tmp_path ) loader = SvnLoader(swh_storage, repo_updated_url, origin_url=repo_initial_url,) assert loader.load() == {"status": "eventful"} visit_status2 = assert_last_visit_matches( loader.storage, repo_updated_url, status="full", type="svn", snapshot=GOURMET_UPDATES_SNAPSHOT.id, ) assert visit_status1.date < visit_status2.date assert visit_status1.snapshot != visit_status2.snapshot stats = get_stats(loader.storage) assert stats == { "content": 22, "directory": 28, "origin": 1, "origin_visit": 2, "release": 0, "revision": 11, "skipped_content": 0, "snapshot": 2, } check_snapshot(GOURMET_UPDATES_SNAPSHOT, loader.storage) # Let's start the ingestion from the start, this should yield the same result loader = SvnLoader( swh_storage, repo_updated_url, origin_url=repo_initial_url, incremental=False, ) assert loader.load() == {"status": "eventful"} visit_status3 = assert_last_visit_matches( loader.storage, repo_updated_url, status="full", type="svn", snapshot=GOURMET_UPDATES_SNAPSHOT.id, ) assert visit_status2.date < visit_status3.date assert visit_status3.snapshot == visit_status2.snapshot check_snapshot(GOURMET_UPDATES_SNAPSHOT, loader.storage) stats = get_stats(loader.storage) assert stats["origin"] == 1 # always the same visit assert stats["origin_visit"] == 2 + 1 # 1 more visit assert stats["snapshot"] == 2 # no new snapshot def test_loader_svn_visit_start_from_revision(swh_storage, datadir, tmp_path): """Starting from existing revision, next visit on changed repo should yield 1 new snapshot. """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_initial_url = prepare_repository_from_archive( archive_path, archive_name, tmp_path ) # repo_initial_url becomes the origin_url we want to visit some more below loader = SvnLoader(swh_storage, repo_initial_url) assert loader.load() == {"status": "eventful"} visit_status1 = assert_last_visit_matches( loader.storage, repo_initial_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) start_revision = loader.storage.revision_get( [hash_to_bytes("95edacc8848369d6fb1608e887d6d2474fd5224f")] )[0] assert start_revision is not None archive_path = os.path.join(datadir, "pkg-gourmet-with-updates.tgz") repo_updated_url = prepare_repository_from_archive( archive_path, "pkg-gourmet", tmp_path ) # we'll start from start_revision - loader = SvnLoader( - swh_storage, - repo_updated_url, - origin_url=repo_initial_url, - swh_revision=start_revision, - ) + loader = SvnLoader(swh_storage, repo_updated_url, origin_url=repo_initial_url,) assert loader.load() == {"status": "eventful"} # nonetheless, we obtain the same snapshot (as previous tests on that repository) visit_status2 = assert_last_visit_matches( loader.storage, repo_updated_url, status="full", type="svn", snapshot=GOURMET_UPDATES_SNAPSHOT.id, ) assert visit_status1.date < visit_status2.date assert visit_status1.snapshot != visit_status2.snapshot stats = get_stats(loader.storage) assert stats == { "content": 22, "directory": 28, "origin": 1, "origin_visit": 2, "release": 0, "revision": 11, "skipped_content": 0, "snapshot": 2, } check_snapshot(GOURMET_UPDATES_SNAPSHOT, loader.storage) def test_loader_svn_visit_with_eol_style(swh_storage, datadir, tmp_path): """Check that a svn repo containing a versioned file with CRLF line endings with svn:eol-style property set to 'native' (this is a violation of svn specification as the file should have been stored with LF line endings) can be loaded anyway. """ archive_name = "mediawiki-repo-r407-eol-native-crlf" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} mediawiki_snapshot = Snapshot( id=hash_to_bytes("d6d6e9703f157c5702d9a4a5dec878926ed4ab76"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("7da4975c363101b819756d33459f30a866d01b1b"), target_type=TargetType.REVISION, ) }, ) check_snapshot(mediawiki_snapshot, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=mediawiki_snapshot.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 1 assert stats["snapshot"] == 1 def test_loader_svn_visit_with_mixed_crlf_lf(swh_storage, datadir, tmp_path): """Check that a svn repo containing a versioned file with mixed CRLF/LF line endings with svn:eol-style property set to 'native' (this is a violation of svn specification as mixed line endings for textual content should not be stored when the svn:eol-style property is set) can be loaded anyway. """ archive_name = "pyang-repo-r343-eol-native-mixed-lf-crlf" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} pyang_snapshot = Snapshot( id=hash_to_bytes("6d9590de11b00a5801de0ff3297c5b44bbbf7d24"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("9c6962eeb9164a636c374be700672355e34a98a7"), target_type=TargetType.REVISION, ) }, ) check_snapshot(pyang_snapshot, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=pyang_snapshot.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 1 assert stats["snapshot"] == 1 def test_loader_svn_with_external_properties(swh_storage, datadir, tmp_path): """Repository with svn:external properties cannot be fully ingested yet """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, "pkg-gourmet-with-external-id.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} gourmet_externals_snapshot = Snapshot( id=hash_to_bytes("19cb68d0a3f22372e2b7017ea5e2a2ea5ae3e09a"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("82a7a4a09f9549223429143ba36ad77375e33c5c"), target_type=TargetType.REVISION, ) }, ) check_snapshot(gourmet_externals_snapshot, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="partial", type="svn", snapshot=gourmet_externals_snapshot.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 1 assert stats["snapshot"] == 1 # repository holds 21 revisions, but the last commit holds an 'svn:externals' # property which will make the loader-svn stops at the last revision prior to the # bad one assert stats["revision"] == 21 - 1 # commit with the svn:external property def test_loader_svn_with_symlink(swh_storage, datadir, tmp_path): """Repository with symlinks should be ingested ok Edge case: - first create a file and commit it. Remove it, then add folder holding the same name, commit. - do the same scenario with symbolic link (instead of file) """ archive_name = "pkg-gourmet" archive_path = os.path.join( datadir, "pkg-gourmet-with-edge-case-links-and-files.tgz" ) repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} gourmet_edge_cases_snapshot = Snapshot( id=hash_to_bytes("18e60982fe521a2546ab8c3c73a535d80462d9d0"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("3f43af2578fccf18b0d4198e48563da7929dc608"), target_type=TargetType.REVISION, ) }, ) check_snapshot(gourmet_edge_cases_snapshot, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=gourmet_edge_cases_snapshot.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 1 assert stats["snapshot"] == 1 assert stats["revision"] == 19 def test_loader_svn_with_wrong_symlinks(swh_storage, datadir, tmp_path): """Repository with wrong symlinks should be ingested ok nonetheless Edge case: - wrong symbolic link - wrong symbolic link with empty space names """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, "pkg-gourmet-with-wrong-link-cases.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} gourmet_wrong_links_snapshot = Snapshot( id=hash_to_bytes("b17f38acabb90f066dedd30c29f01a02af88a5c4"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("cf30d3bb9d5967d0a2bbeacc405f10a5dd9b138a"), target_type=TargetType.REVISION, ) }, ) check_snapshot(gourmet_wrong_links_snapshot, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=gourmet_wrong_links_snapshot.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 1 assert stats["snapshot"] == 1 assert stats["revision"] == 21 def test_loader_svn_loader_from_remote_dump(swh_storage, datadir, tmp_path): """Repository with wrong symlinks should be ingested ok nonetheless Edge case: - wrong symbolic link - wrong symbolic link with empty space names """ archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loaderFromDump = SvnLoaderFromRemoteDump(swh_storage, repo_url) assert loaderFromDump.load() == {"status": "eventful"} assert_last_visit_matches( loaderFromDump.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) origin_url = repo_url + "2" # rename to another origin loader = SvnLoader(swh_storage, repo_url, origin_url=origin_url) assert loader.load() == {"status": "eventful"} # because are working on new origin assert_last_visit_matches( loader.storage, origin_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) stats = get_stats(loader.storage) assert stats["origin"] == 2 # created one more origin assert stats["origin_visit"] == 2 assert stats["snapshot"] == 1 loader = SvnLoader(swh_storage, repo_url) # no change on the origin-url assert loader.load() == {"status": "uneventful"} assert_last_visit_matches( loader.storage, origin_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 2 assert stats["origin_visit"] == 3 assert stats["snapshot"] == 1 # second visit from the dump should be uneventful loaderFromDump = SvnLoaderFromRemoteDump(swh_storage, repo_url) assert loaderFromDump.load() == {"status": "uneventful"} def test_loader_user_defined_svn_properties(swh_storage, datadir, tmp_path): """Edge cases: The repository held some user defined svn-properties with special encodings, this prevented the repository from being loaded even though we do not ingest those information. """ archive_name = "httthttt" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url) assert loader.load() == {"status": "eventful"} expected_snapshot = Snapshot( id=hash_to_bytes("70487267f682c07e52a2371061369b6cf5bffa47"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("604a17dbb15e8d7ecb3e9f3768d09bf493667a93"), target_type=TargetType.REVISION, ) }, ) check_snapshot(expected_snapshot, loader.storage) assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=expected_snapshot.id, ) stats = get_stats(loader.storage) assert stats["origin"] == 1 assert stats["origin_visit"] == 1 assert stats["snapshot"] == 1 assert stats["revision"] == 7 def test_loader_svn_dir_added_then_removed(swh_storage, datadir, tmp_path): """Loader should handle directory removal when processing a commit""" archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}-add-remove-dir.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) - loader = SvnLoader(swh_storage, repo_url, destination_path=tmp_path) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) def test_loader_svn_loader_from_dump_archive(swh_storage, datadir, tmp_path): archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) origin_url = f"svn://{archive_name}" dump_filename = f"{archive_name}.dump" with open(os.path.join(tmp_path, dump_filename), "wb") as dump_file: # create compressed dump file of pkg-gourmet repo subprocess.run(["svnrdump", "dump", repo_url], stdout=dump_file) subprocess.run(["gzip", dump_filename], cwd=tmp_path) # load svn repo from that compressed dump file loader = SvnLoaderFromDumpArchive( swh_storage, url=origin_url, archive_path=os.path.join(tmp_path, f"{dump_filename}.gz"), + temp_directory=tmp_path, ) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, origin_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) assert get_stats(loader.storage) == { "content": 19, "directory": 17, "origin": 1, "origin_visit": 1, "release": 0, "revision": 6, "skipped_content": 0, "snapshot": 1, } class CommitChangeType(Enum): AddOrUpdate = 1 Delete = 2 class CommitChange(TypedDict, total=False): change_type: CommitChangeType path: str properties: Dict[str, str] data: bytes def add_commit(repo_url: str, message: str, changes: List[CommitChange]) -> None: conn = RemoteAccess(repo_url, auth=Auth([get_username_provider()])) editor = conn.get_commit_editor({"svn:log": message}) root = editor.open_root() for change in changes: if change["change_type"] == CommitChangeType.Delete: root.delete_entry(change["path"].rstrip("/")) else: dir_change = change["path"].endswith("/") split_path = change["path"].rstrip("/").split("/") for i in range(len(split_path)): path = "/".join(split_path[0 : i + 1]) if i < len(split_path) - 1: try: root.add_directory(path).close() except SubversionException: pass else: if dir_change: root.add_directory(path).close() else: try: file = root.add_file(path) except SubversionException: file = root.open_file(path) if "properties" in change: for prop, value in change["properties"].items(): file.change_prop(prop, value) if "data" in change: txdelta = file.apply_textdelta() delta.send_stream(BytesIO(change["data"]), txdelta) file.close() root.close() editor.close() def test_loader_eol_style_file_property_handling_edge_case(swh_storage, tmp_path): # create a repository repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" # # first commit add_commit( repo_url, ( "Add a directory containing a file with CRLF end of line " "and set svn:eol-style property to native so CRLF will be " "replaced by LF in the file when exporting the revision" ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="directory/file_with_crlf_eol.txt", properties={"svn:eol-style": "native"}, data=b"Hello world!\r\n", ) ], ) # second commit add_commit( repo_url, "Remove previously added directory and file", [CommitChange(change_type=CommitChangeType.Delete, path="directory/",)], ) # third commit add_commit( repo_url, ( "Add again same directory containing same file with CRLF end of line " "but do not set svn:eol-style property value so CRLF will not be " "replaced by LF when exporting the revision" ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="directory/file_with_crlf_eol.txt", data=b"Hello world!\r\n", ) ], ) # instantiate a svn loader checking after each processed revision that # the repository filesystem it reconstructed does not differ from a subversion # export of that revision - loader = SvnLoader( - swh_storage, repo_url, destination_path=tmp_path, check_revision=1 - ) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path, check_revision=1) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) assert get_stats(loader.storage) == { "content": 2, "directory": 5, "origin": 1, "origin_visit": 1, "release": 0, "revision": 3, "skipped_content": 0, "snapshot": 1, } def get_head_revision_paths_info(loader: SvnLoader) -> Dict[bytes, Dict[str, Any]]: assert loader.snapshot is not None root_dir = loader.snapshot.branches[b"HEAD"].target revision = loader.storage.revision_get([root_dir])[0] assert revision is not None paths = {} for entry in loader.storage.directory_ls(revision.directory, recursive=True): paths[entry["name"]] = entry return paths def test_loader_eol_style_on_svn_link_handling(swh_storage, tmp_path): # create a repository repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" # first commit add_commit( repo_url, ( "Add a regular file, a directory and a link to the regular file " "in the directory. Set svn:eol-style property for the regular " "file and the link. Set svn:special property for the link." ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="file_with_crlf_eol.txt", properties={"svn:eol-style": "native"}, data=b"Hello world!\r\n", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="directory/file_with_crlf_eol.txt", properties={"svn:eol-style": "native", "svn:special": "*"}, data=b"link ../file_with_crlf_eol.txt", ), ], ) # instantiate a svn loader checking after each processed revision that # the repository filesystem it reconstructed does not differ from a subversion # export of that revision - loader = SvnLoader( - swh_storage, repo_url, destination_path=tmp_path, check_revision=1 - ) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path, check_revision=1) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) # check loaded objects are those expected assert get_stats(loader.storage) == { "content": 2, "directory": 2, "origin": 1, "origin_visit": 1, "release": 0, "revision": 1, "skipped_content": 0, "snapshot": 1, } paths = get_head_revision_paths_info(loader) assert ( loader.storage.content_get_data(paths[b"file_with_crlf_eol.txt"]["sha1"]) == b"Hello world!\n" ) assert paths[b"directory/file_with_crlf_eol.txt"]["perms"] == DentryPerms.symlink assert ( loader.storage.content_get_data( paths[b"directory/file_with_crlf_eol.txt"]["sha1"] ) == b"../file_with_crlf_eol.txt" ) def test_loader_svn_special_property_unset(swh_storage, tmp_path): # create a repository repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" # first commit add_commit( repo_url, ( "Create a regular file, a link to a file and a link to an " "external file. Set the svn:special property on the links." ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="file.txt", data=b"Hello world!\n", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="link.txt", properties={"svn:special": "*"}, data=b"link ./file.txt", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="external_link.txt", properties={"svn:special": "*"}, data=b"link /home/user/data.txt", ), ], ) # second commit add_commit( repo_url, "Unset the svn:special property on the links.", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="link.txt", properties={"svn:special": None}, ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="external_link.txt", properties={"svn:special": None}, ), ], ) # instantiate a svn loader checking after each processed revision that # the repository filesystem it reconstructed does not differ from a subversion # export of that revision - loader = SvnLoader( - swh_storage, repo_url, destination_path=tmp_path, check_revision=1 - ) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path, check_revision=1) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) # check loaded objects are those expected assert get_stats(loader.storage) == { "content": 5, "directory": 2, "origin": 1, "origin_visit": 1, "release": 0, "revision": 2, "skipped_content": 0, "snapshot": 1, } paths = get_head_revision_paths_info(loader) assert paths[b"link.txt"]["perms"] == DentryPerms.content assert ( loader.storage.content_get_data(paths[b"link.txt"]["sha1"]) == b"link ./file.txt" ) assert paths[b"external_link.txt"]["perms"] == DentryPerms.content assert ( loader.storage.content_get_data(paths[b"external_link.txt"]["sha1"]) == b"link /home/user/data.txt" ) def test_loader_invalid_svn_eol_style_property_value(swh_storage, tmp_path): # create a repository repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" filename = "file_with_crlf_eol.txt" file_content = b"Hello world!\r\n" # # first commit add_commit( repo_url, ( "Add a file with CRLF end of line and set svn:eol-style property " "to an invalid value." ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path=filename, properties={"svn:eol-style": "foo"}, data=file_content, ) ], ) # instantiate a svn loader checking after each processed revision that # the repository filesystem it reconstructed does not differ from a subversion # export of that revision - loader = SvnLoader( - swh_storage, repo_url, destination_path=tmp_path, check_revision=1 - ) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path, check_revision=1) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) paths = get_head_revision_paths_info(loader) # end of lines should not have been processed assert ( loader.storage.content_get_data(paths[filename.encode()]["sha1"]) == file_content ) def test_loader_first_revision_is_not_number_one(swh_storage, mocker, tmp_path): class SvnRepoSkipFirstRevision(SvnRepo): def logs(self, revision_start, revision_end): """Overrides logs method to skip revision number one in yielded revisions""" yield from super().logs(revision_start + 1, revision_end) from swh.loader.svn import loader mocker.patch.object(loader, "SvnRepo", SvnRepoSkipFirstRevision) # create a repository repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" for filename in ("foo", "bar", "baz"): add_commit( repo_url, f"Add {filename} file", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path=filename, data=f"{filename}\n".encode(), ) ], ) - loader = SvnLoader(swh_storage, repo_url, destination_path=tmp_path) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path) # post loading will detect an issue and make a partial visit with a snapshot assert loader.load() == {"status": "failed"} assert_last_visit_matches( loader.storage, repo_url, status="partial", type="svn", ) assert get_stats(loader.storage) == { "content": 2, "directory": 2, "origin": 1, "origin_visit": 1, "release": 0, "revision": 2, "skipped_content": 0, "snapshot": 1, } def test_loader_svn_special_property_on_binary_file_with_null_byte( swh_storage, tmp_path ): """When a file has the svn:special property set but is not a svn link, it will be truncated when performing an export operation if it contains a null byte. Indeed, subversion will treat the file content as text but it might be a binary file containing null bytes.""" # create a repository repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" data = ( b"!\xff\xfea\x00p\x00t\x00-\x00c\x00y\x00g\x00.\x00s\x00h\x00\x00\x00" ) # first commit add_commit( repo_url, "Add a non svn link binary file and set the svn:special property on it", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="binary_file", properties={"svn:special": "*"}, data=data, ), ], ) # second commit add_commit( repo_url, "Remove the svn:special property on the previously added file", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="binary_file", properties={"svn:special": None}, ), ], ) # instantiate a svn loader checking after each processed revision that # the repository filesystem it reconstructed does not differ from a subversion # export of that revision - loader = SvnLoader( - swh_storage, repo_url, destination_path=tmp_path, check_revision=1 - ) + loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path, check_revision=1) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) def test_loader_last_revision_divergence(swh_storage, datadir, tmp_path): archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) class SvnLoaderRevisionDivergence(SvnLoader): def _check_revision_divergence(self, count, rev, dir_id): raise ValueError("revision divergence detected") - loader = SvnLoaderRevisionDivergence( - swh_storage, repo_url, destination_path=tmp_path - ) + loader = SvnLoaderRevisionDivergence(swh_storage, repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "failed"} assert_last_visit_matches( loader.storage, repo_url, status="partial", type="svn", snapshot=GOURMET_SNAPSHOT.id, )