diff --git a/swh/loader/svn/loader.py b/swh/loader/svn/loader.py index 2d898ed..6af4ebe 100644 --- a/swh/loader/svn/loader.py +++ b/swh/loader/svn/loader.py @@ -1,820 +1,826 @@ # 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 Any, 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, 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): """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, incremental: bool = True, temp_directory: str = "/tmp", debug: bool = False, check_revision: int = 0, **kwargs: Any, ): """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 """ # technical svn uri to act on svn repository self.svn_url = url # origin url as unique identifier for origin in swh archive origin_url = origin_url or self.svn_url super().__init__(storage=storage, origin_url=origin_url, **kwargs) self.debug = debug self.temp_directory = temp_directory self.done = False self.svnrepo = None self.skip_post_load = False # 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 or self.visit_date self.incremental = incremental self.snapshot: Optional[Snapshot] = None # state from previous visit self.latest_snapshot = None self.latest_revision: Optional[Revision] = None self.from_dump = False 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: int) -> bytes: """Compute and return the hash tree at a given svn revision. Args: 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) root_dir = from_disk.Directory.from_disk(path=local_url) self.svnrepo.clean_fs(local_dirname) return root_dir.hash 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: 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: 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, 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 assert self.svnrepo is not None commit, root_dir = self.svnrepo.swh_hash_data_at_revision(revision_start) dir_id = root_dir.hash swh_revision = self.build_swh_revision(revision_start, commit, dir_id, parents) swh_revision_id = swh_revision.id return swh_revision_id == revision_id def start_from(self) -> Tuple[int, int]: """Determine from where to start the loading. Returns: tuple (revision_start, revision_end) 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 # 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"]) if not self.check_history_not_altered(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 = 0 # now we know history is ok, we start at next revision revision_start = revision_start + 1 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 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 regular check will help spotting potential missing edge cases. Args: 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 """ 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 ) -> 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) parents = (self.latest_revision.id,) if self.latest_revision is not None else () count = 0 for rev, 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, parents) 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 and self.check_revision != 0 and count % self.check_revision == 0 ): self._check_revision_divergence(rev, dir_id) parents = (swh_revision.id,) yield _contents, _skipped_contents, _directories, swh_revision if not self.debug and self.svnrepo: # clean directory where revisions were replayed to gain some disk space # before the post_load operation self.svnrepo.clean_fs(self.svnrepo.local_url) def svn_repo(self, *args, **kwargs): """Wraps the creation of SvnRepo object and handles not found repository errors.""" try: return SvnRepo(*args, **kwargs) except SubversionException as e: error_msgs = [ "Unable to connect to a repository at URL", "Unknown URL type", + "is not a working copy", ] for msg in error_msgs: if msg in e.args[0]: self._load_status = "uneventful" raise NotFound(e) raise def prepare(self): if self.incremental: latest_snapshot_revision = self._latest_snapshot_revision(self.origin.url) if latest_snapshot_revision: self.latest_snapshot, self.latest_revision = latest_snapshot_revision self._snapshot = self.latest_snapshot self._last_revision = self.latest_revision local_dirname = self._create_tmp_dir(self.temp_directory) self.svnrepo = self.svn_repo( self.svn_url, self.origin.url, local_dirname, self.max_content_size, self.from_dump, ) try: revision_start, revision_end = self.start_from() self.swh_revision_gen = self.process_svn_revisions( self.svnrepo, revision_start, revision_end ) except SvnLoaderUneventful as e: self.log.warning(e) 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. """ if self.done: return False try: data = next(self.swh_revision_gen) self._load_status = "eventful" except StopIteration: self.done = True # Stopping iteration self._visit_status = "full" except Exception as e: # svn:external, hash divergence, i/o error... self.log.exception(e) self.done = True # Stopping iteration self._visit_status = "partial" else: self._contents, self._skipped_contents, self._directories, rev = data if rev: self._last_revision = rev self._revisions.append(rev) return not self.done 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 self.skip_post_load: return if success and self._last_revision is not None: # 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 # the case, an exception will be raised to report the issue and mark the # visit as partial self._check_revision_divergence( int(dict(self._last_revision.extra_headers)[b"svn_revision"]), self._last_revision.directory, ) def _create_tmp_dir(self, root_tmp_dir: str) -> str: return tempfile.mkdtemp( dir=root_tmp_dir, prefix=TEMPORARY_DIR_PREFIX_PATTERN, suffix="-%s" % os.getpid(), ) class SvnLoaderFromDumpArchive(SvnLoader): """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, incremental: bool = False, visit_date: Optional[datetime] = None, temp_directory: str = "/tmp", debug: bool = False, check_revision: int = 0, **kwargs: Any, ): super().__init__( storage=storage, url=url, origin_url=origin_url, incremental=incremental, visit_date=visit_date, temp_directory=temp_directory, debug=debug, check_revision=check_revision, **kwargs, ) self.archive_path = archive_path self.temp_dir = None self.repo_path = None self.from_dump = True 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 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, incremental: bool = True, visit_date: Optional[datetime] = None, temp_directory: str = "/tmp", debug: bool = False, check_revision: int = 0, **kwargs: Any, ): super().__init__( storage=storage, url=url, origin_url=origin_url, incremental=incremental, visit_date=visit_date, temp_directory=temp_directory, debug=debug, check_revision=check_revision, **kwargs, ) self.from_dump = True self.temp_dir = self._create_tmp_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: 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 for stderr_line in reversed(stderr_lines): if stderr_line.startswith("* Dumped revision"): last_dumped_rev = stderr_line break 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 and last_dumped_rev < last_loaded_svn_rev: 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) + last_loaded_svn_rev = self.get_last_loaded_svn_rev(self.origin.url) + + self.svnrepo = self.svn_repo( + self.origin.url, self.origin.url, self.temp_dir, self.max_content_size + ) + + # Ensure to use remote URL retrieved by SvnRepo as origin URL might redirect + # and svnrdump does not handle URL redirection + self.svn_url = self.svnrepo.remote_url # Then for stale repository, check if the last loaded revision in the archive # is different from the last revision on the remote subversion server. # Skip the dump of all revisions and the loading process if they are identical # to save some disk space and processing time. last_loaded_snp_and_rev = self._latest_snapshot_revision(self.origin.url) if last_loaded_snp_and_rev is not None: last_loaded_snp, last_loaded_rev = last_loaded_snp_and_rev - self.svnrepo = self.svn_repo( - self.origin.url, self.origin.url, self.temp_dir, self.max_content_size - ) stale_repository = self.svnrepo.head_revision() == last_loaded_svn_rev if stale_repository and self.check_history_not_altered( last_loaded_svn_rev, last_loaded_rev ): self._snapshot = last_loaded_snp self._last_revision = last_loaded_rev self.done = True self.skip_post_load = True return # 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 984f9d9..079461c 100644 --- a/swh/loader/svn/svn.py +++ b/swh/loader/svn/svn.py @@ -1,535 +1,540 @@ # Copyright (C) 2015-2022 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 typing import Dict, Iterator, List, Optional, Tuple, Union from subvertpy import SubversionException, client, properties, wc from subvertpy.ra import Auth, RemoteAccess, get_username_provider from swh.model.from_disk import Directory as DirectoryFromDisk from swh.model.model import ( Content, Directory, Person, SkippedContent, TimestampWithTimezone, ) from . import converters, replay from .svn_retry import svn_retry from .utils import is_recursive_external, parse_external_definition # When log message contains empty data DEFAULT_AUTHOR_MESSAGE = "" logger = logging.getLogger(__name__) class SvnRepo: """Svn repository representation. Args: 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: str, origin_url: str, local_dirname: str, max_content_length: int, from_dump: bool = False, ): self.remote_url = remote_url.rstrip("/") self.origin_url = origin_url self.from_dump = from_dump auth = Auth([get_username_provider()]) + # one client for update operation + self.client = client.Client(auth=auth) + + if not self.remote_url.startswith("file://"): + # use redirection URL if any for remote operations + self.remote_url = self.info(self.remote_url).url + # one connection for log iteration self.conn_log = self.remote_access(auth) # another for replay self.conn = self.remote_access(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 = replay.Replay( conn=self.conn, rootpath=self.local_url, svnrepo=self, temp_dir=local_dirname, ) self.max_content_length = max_content_length self.has_relative_externals = False self.has_recursive_externals = False self.replay_started = False # compute root directory path from the remote repository URL, required to # properly load the sub-tree of a repository mounted from a dump file repos_root_url = self.info(origin_url).repos_root_url self.root_directory = origin_url.replace(repos_root_url, "", 1) 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) -> int: """Retrieve current head revision.""" return self.conn.get_latest_revnum() def initial_revision(self) -> int: """Retrieve the initial revision from which the remote url appeared.""" return 1 def convert_commit_message(self, msg: Union[str, bytes]) -> bytes: """Simply encode the commit message. Args: msg: 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: bytes) -> TimestampWithTimezone: """Convert the message commit date into a timestamp in swh format. The precision is kept. Args: date: the commit date to convert. Returns: The transformed date. """ return converters.svn_date_to_swh_date(date) def convert_commit_author(self, author: Optional[bytes]) -> Person: """Convert the commit author into an swh person. Args: author: the commit author to convert. Returns: Person as model object """ return converters.svn_author_to_swh_person(author) def __to_entry(self, log_entry: Tuple) -> Dict: 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) ) has_changes = ( not self.from_dump or changed_paths is not None and any( changed_path.startswith(self.root_directory) for changed_path in changed_paths.keys() ) ) return { "rev": rev, "author_date": author_date, "author_name": author, "message": message, "has_changes": has_changes, } def logs(self, revision_start: int, revision_end: int) -> Iterator[Dict]: """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=self.from_dump, ): yield self.__to_entry(log_entry) @svn_retry() def remote_access(self, auth: Auth) -> RemoteAccess: """Simple wrapper around subvertpy.ra.RemoteAccess creation enabling to retry the operation if a network error occurs.""" return RemoteAccess(self.remote_url, auth=auth) @svn_retry() def info(self, origin_url: str): """Simple wrapper around subvertpy.client.Client.info enabling to retry the command if a network error occurs.""" info = self.client.info(origin_url.rstrip("/")) return next(iter(info.values())) @svn_retry() def export( self, url: str, to: str, rev: Optional[int] = None, peg_rev: Optional[int] = None, recurse: bool = True, ignore_externals: bool = False, overwrite: bool = False, ignore_keywords: bool = False, ) -> int: """Simple wrapper around subvertpy.client.Client.export enabling to retry the command if a network error occurs. See documentation of svn_client_export5 function from subversion C API to get details about parameters. """ # remove export path as command can be retried if os.path.isfile(to) or os.path.islink(to): os.remove(to) elif os.path.isdir(to): shutil.rmtree(to) options = [] if rev is not None: options.append(f"-r {rev}") if recurse: options.append("--depth infinity") if ignore_externals: options.append("--ignore-externals") if overwrite: options.append("--force") if ignore_keywords: options.append("--ignore-keywords") logger.debug( "svn export %s %s%s %s", " ".join(options), url, f"@{peg_rev}" if peg_rev else "", to, ) return self.client.export( url, to=to, rev=rev, peg_rev=peg_rev, recurse=recurse, ignore_externals=ignore_externals, overwrite=overwrite, ignore_keywords=ignore_keywords, ) @svn_retry() def checkout( self, url: str, path: str, rev: Optional[int] = None, peg_rev: Optional[int] = None, recurse: bool = True, ignore_externals: bool = False, allow_unver_obstructions: bool = False, ) -> int: """Simple wrapper around subvertpy.client.Client.checkout enabling to retry the command if a network error occurs. See documentation of svn_client_checkout3 function from subversion C API to get details about parameters. """ if os.path.isdir(os.path.join(path, ".svn")): # cleanup checkout path as command can be retried and svn working copy might # be locked wc.cleanup(path) elif os.path.isdir(path): # recursively remove checkout path otherwise if it is not a svn working copy shutil.rmtree(path) options = [] if rev is not None: options.append(f"-r {rev}") if recurse: options.append("--depth infinity") if ignore_externals: options.append("--ignore-externals") logger.debug( "svn checkout %s %s%s %s", " ".join(options), self.remote_url, f"@{peg_rev}" if peg_rev else "", path, ) return self.client.checkout( url, path=path, rev=rev, peg_rev=peg_rev, recurse=recurse, ignore_externals=ignore_externals, allow_unver_obstructions=allow_unver_obstructions, ) @svn_retry() def propget( self, name: str, target: str, peg_rev: Optional[int], rev: Optional[int] = None, recurse: bool = False, ): """Simple wrapper around subvertpy.client.Client.propget enabling to retry the command if a network error occurs. See documentation of svn_client_propget5 function from subversion C API to get details about parameters. """ return self.client.propget(name, target, peg_rev, rev, recurse) def export_temporary(self, revision: int) -> Tuple[str, bytes]: """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( dir=self.local_dirname, prefix=f"check-revision-{revision}." ) local_name = os.path.basename(self.remote_url) local_url = os.path.join(local_dirname, local_name) url = self.remote_url # if some paths have external URLs relative to the repository URL but targeting # paths outside it, we need to export from the origin URL as the remote URL can # target a dump mounted on the local filesystem if self.replay_started and self.has_relative_externals: # externals detected while replaying revisions url = self.origin_url elif not self.replay_started: # revisions replay has not started, we need to check if svn:externals # properties are set from a checkout of the revision and if some # external URLs are relative to pick the right export URL, # recursive externals are also checked with tempfile.TemporaryDirectory( dir=self.local_dirname, prefix=f"checkout-revision-{revision}." ) as co_dirname: self.checkout( self.remote_url, co_dirname, revision, ignore_externals=True ) # get all svn:externals properties recursively externals = self.propget("svn:externals", co_dirname, None, None, True) self.has_relative_externals = False self.has_recursive_externals = False for path, external_defs in externals.items(): if self.has_relative_externals or self.has_recursive_externals: break path = path.replace(self.remote_url.rstrip("/") + "/", "") for external_def in os.fsdecode(external_defs).split("\n"): # skip empty line or comment if not external_def or external_def.startswith("#"): continue ( external_path, external_url, _, relative_url, ) = parse_external_definition( external_def.rstrip("\r"), path, self.origin_url ) if is_recursive_external( self.origin_url, path, external_path, external_url, ): self.has_recursive_externals = True url = self.remote_url break if relative_url: self.has_relative_externals = True url = self.origin_url break try: url = url.rstrip("/") self.export( url, to=local_url, rev=revision, ignore_keywords=True, ignore_externals=self.has_recursive_externals, ) except SubversionException as se: if se.args[0].startswith( ( "Error parsing svn:externals property", "Unrecognized format for the relative external URL", ) ): pass else: raise if self.from_dump: # when exporting a subpath of a subversion repository mounted from # a dump file generated by svnrdump, exported paths are relative to # the repository root path while they are relative to the subpath # otherwise, so we need to adjust the URL of the exported filesystem root_dir_local_url = os.path.join(local_url, self.root_directory.strip("/")) # check that root directory of a subproject did not get removed in revision if os.path.exists(root_dir_local_url): local_url = root_dir_local_url return local_dirname, os.fsencode(local_url) def swh_hash_data_per_revision( self, start_revision: int, end_revision: int ) -> Iterator[ Tuple[ int, Dict, Tuple[List[Content], List[SkippedContent], List[Directory]], DirectoryFromDisk, ], ]: """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 - commit: commit data (author, date, message) for such revision - objects_per_path: Tuple of list of objects between start_revision and end_revision - complete Directory representation """ # even in incremental loading mode, we need to replay the whole set of # path modifications from first revision to restore possible file states induced # by setting svn properties on those files (end of line style for instance) self.replay_started = True first_revision = 1 if start_revision else 0 # handle empty repository edge case for commit in self.logs(first_revision, end_revision): rev = commit["rev"] objects = self.swhreplay.compute_objects(rev) if rev >= start_revision: # start yielding new data to archive once we reached the revision to # resume the loading from if commit["has_changes"] or start_revision == 0: # yield data only if commit has changes or if repository is empty root_dir_path = self.root_directory.encode()[1:] if not root_dir_path or root_dir_path in self.swhreplay.directory: root_dir = self.swhreplay.directory[root_dir_path] else: # root directory of subproject got removed in revision, return # empty directory for that edge case root_dir = DirectoryFromDisk() yield rev, commit, objects, root_dir def swh_hash_data_at_revision( self, revision: int ) -> Tuple[Dict, DirectoryFromDisk]: """Compute the information at a given svn revision. This is expected to be used for checks only. Yields: The tuple (commit dictionary, targeted directory object). """ # Update disk representation of the repository at revision id local_dirname, local_url = self.export_temporary(revision) # Compute the current hashes on disk directory = DirectoryFromDisk.from_disk( path=local_url, max_content_length=self.max_content_length ) # Retrieve the commit information for revision commit = list(self.logs(revision, revision))[0] # Clean export directory self.clean_fs(local_dirname) return commit, directory def clean_fs(self, local_dirname: Optional[str] = None) -> None: """Clean up the local working copy. Args: local_dirname: Path to remove recursively if provided. Otherwise, remove the temporary upper root tree used for svn repository loading. """ dirname = local_dirname or self.local_dirname if os.path.exists(dirname): logger.debug("cleanup %s", dirname) shutil.rmtree(dirname) diff --git a/swh/loader/svn/tests/test_loader.py b/swh/loader/svn/tests/test_loader.py index 43850c7..4fca79f 100644 --- a/swh/loader/svn/tests/test_loader.py +++ b/swh/loader/svn/tests/test_loader.py @@ -1,2067 +1,2100 @@ # Copyright (C) 2016-2022 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 import os import shutil import subprocess from typing import Any, Dict import pytest from subvertpy import SubversionException 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 from .utils import CommitChange, CommitChangeType, add_commit 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, 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, 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, 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, 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, temp_directory=loading_path) assert loader.load() == {"status": "eventful"} actual_visit = assert_last_visit_matches( swh_storage, repo_url, status="full", type="svn", ) check_snapshot(loader.snapshot, loader.storage) loader2 = SvnLoaderFromRemoteDump( 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, 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, ) check_snapshot(loader.snapshot, loader.storage) 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, temp_directory=tmp_path) assert loader.load() == {"status": "eventful"} visit_status1 = assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(loader.snapshot, loader.storage) 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, temp_directory=tmp_path) 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, temp_directory=tmp_path) 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, ) check_snapshot(loader.snapshot, loader.storage) 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, temp_directory=tmp_path ) assert loader2.load() == {"status": "eventful"} assert_last_visit_matches( loader2.storage, repo_url, status="full", type="svn", snapshot=hash_to_bytes("5aa61959e788e281fd6e187053d0f46c68e8d8bb"), ) check_snapshot(loader.snapshot, loader.storage) 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, temp_directory=tmp_path) assert loader.load() == {"status": "eventful"} visit_status1 = assert_last_visit_matches( loader.storage, repo_initial_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) 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, temp_directory=tmp_path, ) 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, temp_directory=tmp_path, ) 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, temp_directory=tmp_path) assert loader.load() == {"status": "eventful"} visit_status1 = assert_last_visit_matches( loader.storage, repo_initial_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) 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, temp_directory=tmp_path, ) 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, temp_directory=tmp_path) 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, temp_directory=tmp_path) 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_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, temp_directory=tmp_path) 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, temp_directory=tmp_path) 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_cleanup_loader(swh_storage, datadir, tmp_path): """Loader should clean up its working directory after the load""" 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) loading_temp_directory = str(tmp_path / "loading") os.mkdir(loading_temp_directory) loader = SvnLoader(swh_storage, repo_url, temp_directory=loading_temp_directory) assert loader.load() == {"status": "eventful"} # the root temporary directory still exists assert os.path.exists(loader.temp_directory) # but it should be empty assert os.listdir(loader.temp_directory) == [] def test_loader_svn_cleanup_loader_from_remote_dump(swh_storage, datadir, tmp_path): """Loader should clean up its working directory after the load""" 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) loading_temp_directory = str(tmp_path / "loading") os.mkdir(loading_temp_directory) loader = SvnLoaderFromRemoteDump( swh_storage, repo_url, temp_directory=loading_temp_directory ) assert loader.load() == {"status": "eventful"} # the root temporary directory still exists assert os.path.exists(loader.temp_directory) # but it should be empty assert os.listdir(loader.temp_directory) == [] # the internal temp_dir should be cleaned up though assert not os.path.exists(loader.temp_dir) def test_loader_svn_cleanup_loader_from_dump_archive(swh_storage, datadir, tmp_path): """Loader should clean up its working directory after the load""" 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, temp_directory=loading_path) assert loader.load() == {"status": "eventful"} # the root temporary directory still exists assert os.path.exists(loader.temp_directory) # but it should be empty assert os.listdir(loader.temp_directory) == [] # the internal temp_dir should be cleaned up though assert not os.path.exists(loader.temp_dir) def test_svn_loader_from_remote_dump(swh_storage, datadir, tmpdir_factory): archive_name = "pkg-gourmet" archive_path = os.path.join(datadir, f"{archive_name}.tgz") tmp_path = tmpdir_factory.mktemp("repo1") repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loaderFromDump = SvnLoaderFromRemoteDump( swh_storage, repo_url, temp_directory=tmp_path ) assert loaderFromDump.load() == {"status": "eventful"} assert_last_visit_matches( loaderFromDump.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) # rename to another origin tmp_path = tmpdir_factory.mktemp("repo2") origin_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader( swh_storage, repo_url, origin_url=origin_url, temp_directory=tmp_path ) 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, temp_directory=tmp_path ) # 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, temp_directory=tmp_path ) assert loaderFromDump.load() == {"status": "uneventful"} def test_svn_loader_from_remote_dump_incremental_load_on_stale_repo( swh_storage, datadir, tmp_path, mocker ): 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) # first load: a dump file will be created, mounted to a local repository # and the latter will be loaded into the archive loaderFromDump = SvnLoaderFromRemoteDump( swh_storage, repo_url, temp_directory=tmp_path ) assert loaderFromDump.load() == {"status": "eventful"} assert_last_visit_matches( loaderFromDump.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) # second load on same repository: the loader will detect there is no changes # since last load and will skip the dump, mount and load phases loaderFromDump = SvnLoaderFromRemoteDump( swh_storage, repo_url, temp_directory=tmp_path ) loaderFromDump.dump_svn_revisions = mocker.MagicMock() init_svn_repo_from_dump = mocker.patch( "swh.loader.svn.loader.init_svn_repo_from_dump" ) loaderFromDump.process_svn_revisions = mocker.MagicMock() loaderFromDump._check_revision_divergence = mocker.MagicMock() assert loaderFromDump.load() == {"status": "uneventful"} assert_last_visit_matches( loaderFromDump.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) # no dump loaderFromDump.dump_svn_revisions.assert_not_called() # no mount init_svn_repo_from_dump.assert_not_called() # no loading loaderFromDump.process_svn_revisions.assert_not_called() # no redundant post_load processing loaderFromDump._check_revision_divergence.assert_not_called() def test_svn_loader_from_remote_dump_incremental_load_on_non_stale_repo( swh_storage, datadir, tmp_path, mocker ): 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) # first load loader = SvnLoaderFromRemoteDump(swh_storage, repo_url, temp_directory=tmp_path) loader.load() archive_path = os.path.join(datadir, "pkg-gourmet-with-updates.tgz") repo_updated_url = prepare_repository_from_archive( archive_path, archive_name, tmp_path ) # second load loader = SvnLoaderFromRemoteDump( swh_storage, repo_updated_url, temp_directory=tmp_path ) dump_svn_revisions = mocker.spy(loader, "dump_svn_revisions") process_svn_revisions = mocker.spy(loader, "process_svn_revisions") loader.load() dump_svn_revisions.assert_called() process_svn_revisions.assert_called() 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, temp_directory=tmp_path) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", ) check_snapshot(loader.snapshot, loader.storage) 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) 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=repo_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, repo_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, } def test_loader_eol_style_file_property_handling_edge_case( swh_storage, repo_url, tmp_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, 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_snapshot(loader.snapshot, loader.storage) 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, repo_url, tmp_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, 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_snapshot(loader.snapshot, loader.storage) # 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, repo_url, tmp_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, 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_snapshot(loader.snapshot, loader.storage) # 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, repo_url, tmp_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, 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_snapshot(loader.snapshot, loader.storage) 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, repo_url, 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) 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, 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", ) check_snapshot(loader.snapshot, loader.storage) 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(swh_storage, repo_url, tmp_path): """When a file has the svn:special property set but is not a svn link, it might be truncated under certain conditions when performing an export operation.""" 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." "That file will be truncated when exporting it." ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="binary_file", properties={"svn:special": "*"}, data=data, ), ], ) # second commit add_commit( repo_url, ( "Add a non svn link binary file and set the svn:special and " "svn:mime-type properties on it." "That file will not be truncated when exporting it." ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="another_binary_file", properties={ "svn:special": "*", "svn:mime-type": "application/octet-stream", }, data=data, ), ], ) # third commit add_commit( repo_url, "Remove the svn:special property on the previously added files", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="binary_file", properties={"svn:special": None}, ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="another_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, 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_snapshot(loader.snapshot, loader.storage) 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, 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, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) def test_loader_delete_directory_while_file_has_same_prefix( swh_storage, repo_url, tmp_path ): # first commit add_commit( repo_url, "Add a file and a directory with same prefix", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="foo/bar.c", data=b'#include "../foo.c"', ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="foo.c", data=b"int foo() {return 0;}", ), ], ) # second commit add_commit( repo_url, "Delete previously added directory and update file content", [ CommitChange(change_type=CommitChangeType.Delete, path="foo"), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="foo.c", data=b"int foo() {return 1;}", ), ], ) # 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, 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_snapshot(loader.snapshot, loader.storage) def test_svn_loader_incremental(swh_storage, repo_url, tmp_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="file_with_crlf_eol.txt", properties={"svn:eol-style": "native"}, data=b"Hello world!\r\n", ) ], ) # first load 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_snapshot(loader.snapshot, loader.storage) # second commit add_commit( repo_url, "Modify previously added file", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="file_with_crlf_eol.txt", data=b"Hello World!\r\n", ) ], ) # second load, incremental 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_snapshot(loader.snapshot, loader.storage) # third commit add_commit( repo_url, "Unset svn:eol-style property on file", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="file_with_crlf_eol.txt", properties={"svn:eol-style": None}, ) ], ) # third load, incremental 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_snapshot(loader.snapshot, loader.storage) def test_svn_loader_incremental_replay_start_with_empty_directory( swh_storage, mocker, repo_url, tmp_path ): # first commit add_commit( repo_url, ("Add a file"), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="foo.txt", data=b"foo\n", ) ], ) # first load 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", ) check_snapshot(loader.snapshot, loader.storage) # second commit add_commit( repo_url, "Modify previously added file", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="foo.txt", data=b"bar\n", ) ], ) class SvnRepoCheckReplayStartWithEmptyDirectory(SvnRepo): def swh_hash_data_per_revision(self, start_revision: int, end_revision: int): """Overrides swh_hash_data_per_revision method to grab the content of the directory where the svn revisions will be replayed before that process starts.""" self.replay_dir_content_before_start = [ os.path.join(root, name) for root, _, files in os.walk(self.local_url) for name in files ] yield from super().swh_hash_data_per_revision(start_revision, end_revision) from swh.loader.svn import loader mocker.patch.object(loader, "SvnRepo", SvnRepoCheckReplayStartWithEmptyDirectory) # second load, incremental loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path) loader.load() # check work directory was empty before replaying revisions assert loader.svnrepo.replay_dir_content_before_start == [] def test_loader_svn_executable_property_on_svn_link_handling( swh_storage, repo_url, tmp_path ): # first commit add_commit( repo_url, ( "Add an executable file and a svn link to it." "Set svn:executable property for both paths." ), [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello-world", properties={"svn:executable": "*"}, data=b"#!/bin/bash\necho Hello World !", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:executable": "*", "svn:special": "*"}, data=b"link hello-world", ), ], ) # second commit add_commit( repo_url, ( "Remove executable file, unset link and replace it with executable content." "As the link was previously marked as executable, execution rights should" "be set after turning it to a regular file." ), [ CommitChange(change_type=CommitChangeType.Delete, path="hello-world"), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:special": None}, data=b"#!/bin/bash\necho Hello World !", ), ], ) # 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, 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_snapshot(loader.snapshot, loader.storage) def test_loader_svn_add_property_on_link(swh_storage, repo_url, tmp_path): # first commit add_commit( repo_url, "Add an executable file and a svn link to it.", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello-world", properties={"svn:executable": "*"}, data=b"#!/bin/bash\necho Hello World !", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:special": "*"}, data=b"link hello-world", ), ], ) # second commit add_commit( repo_url, "Set svn:eol-style property on link", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:eol-style": "native"}, ), ], ) # 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, 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_snapshot(loader.snapshot, loader.storage) def test_loader_svn_link_parsing(swh_storage, repo_url, tmp_path): # first commit add_commit( repo_url, "Add an executable file and a svn link to it.", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello-world", properties={"svn:executable": "*"}, data=b"#!/bin/bash\necho Hello World !", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:special": "*"}, data=b"link hello-world", ), ], ) # second commit add_commit( repo_url, "Update svn link content", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", data=b"link 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, 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_snapshot(loader.snapshot, loader.storage) def test_loader_svn_empty_local_dir_before_post_load(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 SvnLoaderPostLoadLocalDirIsEmpty(SvnLoader): def post_load(self, success=True): if success: self.local_dirname_content = [ os.path.join(root, name) for root, _, files in os.walk(self.svnrepo.local_dirname) for name in files ] return super().post_load(success) loader = SvnLoaderPostLoadLocalDirIsEmpty( swh_storage, repo_url, temp_directory=tmp_path ) assert loader.load() == {"status": "eventful"} assert loader.local_dirname_content == [] assert_last_visit_matches( loader.storage, repo_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) def _dump_project(tmp_path, origin_url): svnrdump_cmd = ["svnrdump", "dump", origin_url] dump_path = f"{tmp_path}/repo.dump" with open(dump_path, "wb") as dump_file: subprocess.run(svnrdump_cmd, stdout=dump_file) subprocess.run(["gzip", dump_path]) return dump_path + ".gz" def test_loader_svn_add_property_on_directory_link(swh_storage, repo_url, tmp_path): # first commit add_commit( repo_url, "Add an executable file in a directory and a svn link to the directory.", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="code/hello-world", properties={"svn:executable": "*"}, data=b"#!/bin/bash\necho Hello World !", ), CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:special": "*"}, data=b"link code", ), ], ) # second commit add_commit( repo_url, "Set svn:eol-style property on link", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="hello", properties={"svn:eol-style": "native"}, ), ], ) # 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, 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_snapshot(loader.snapshot, loader.storage) @pytest.mark.parametrize( "svn_loader_cls", [SvnLoader, SvnLoaderFromDumpArchive, SvnLoaderFromRemoteDump] ) def test_loader_with_subprojects(swh_storage, repo_url, tmp_path, svn_loader_cls): # first commit add_commit( repo_url, "Add first project in repository", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="project1/foo.sh", data=b"#!/bin/bash\necho foo", ), ], ) # second commit add_commit( repo_url, "Add second project in repository", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="project2/bar.sh", data=b"#!/bin/bash\necho bar", ), ], ) # third commit add_commit( repo_url, "Add third project in repository", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="project3/baz.sh", data=b"#!/bin/bash\necho baz", ), ], ) for i in range(1, 4): # load each project in the repository separately origin_url = f"{repo_url}/project{i}" loader_params = { "storage": swh_storage, "url": origin_url, "origin_url": origin_url, "temp_directory": tmp_path, "incremental": True, "check_revision": 1, } if svn_loader_cls == SvnLoaderFromDumpArchive: loader_params["archive_path"] = _dump_project(tmp_path, origin_url) loader = svn_loader_cls(**loader_params) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, origin_url, status="full", type="svn", ) check_snapshot(loader.snapshot, loader.storage) if svn_loader_cls == SvnLoaderFromDumpArchive: loader_params["archive_path"] = _dump_project(tmp_path, origin_url) loader = svn_loader_cls(**loader_params) assert loader.load() == {"status": "uneventful"} # each project origin must have assert get_stats(loader.storage) == { "content": i, # one content "directory": 2 * i, # two directories "origin": i, "origin_visit": 2 * i, # two visits "release": 0, "revision": i, # one revision "skipped_content": 0, "snapshot": i, # one snapshot } @pytest.mark.parametrize( "svn_loader_cls", [SvnLoader, SvnLoaderFromDumpArchive, SvnLoaderFromRemoteDump] ) def test_loader_subproject_root_dir_removal( swh_storage, repo_url, tmp_path, svn_loader_cls ): # first commit add_commit( repo_url, "Add project in repository", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="project/foo.sh", data=b"#!/bin/bash\necho foo", ), ], ) # second commit add_commit( repo_url, "Remove project root directory", [CommitChange(change_type=CommitChangeType.Delete, path="project/")], ) # third commit add_commit( repo_url, "Re-add project in repository", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path="project/foo.sh", data=b"#!/bin/bash\necho foo", ), ], ) origin_url = f"{repo_url}/project" loader_params = { "storage": swh_storage, "url": origin_url, "origin_url": origin_url, "temp_directory": tmp_path, "incremental": True, "check_revision": 1, } if svn_loader_cls == SvnLoaderFromDumpArchive: loader_params["archive_path"] = _dump_project(tmp_path, origin_url) loader = svn_loader_cls(**loader_params) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, origin_url, status="full", type="svn", ) check_snapshot(loader.snapshot, loader.storage) if svn_loader_cls == SvnLoaderFromDumpArchive: loader_params["archive_path"] = _dump_project(tmp_path, origin_url) loader = svn_loader_cls(**loader_params) assert loader.load() == {"status": "uneventful"} @pytest.mark.parametrize("svn_loader_cls", [SvnLoader, SvnLoaderFromRemoteDump]) def test_loader_svn_not_found_after_successful_visit( swh_storage, datadir, tmp_path, svn_loader_cls ): 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 = svn_loader_cls(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, ) check_snapshot(loader.snapshot, loader.storage) # simulate removal of remote repository shutil.rmtree(repo_url.replace("file://", "")) loader = svn_loader_cls(swh_storage, repo_url, temp_directory=tmp_path) assert loader.load() == {"status": "uneventful"} assert_last_visit_matches( loader.storage, repo_url, status="not_found", type="svn", snapshot=None, ) + + +def test_loader_svn_from_remote_dump_url_redirect(swh_storage, tmp_path, mocker): + repo_url = "http://svn.example.org/repo" + repo_redirect_url = "https://svn.example.org/repo" + + # mock remote subversion operations + from swh.loader.svn.svn import client + + mocker.patch("swh.loader.svn.svn.RemoteAccess") + init_svn_repo_from_dump = mocker.patch( + "swh.loader.svn.loader.init_svn_repo_from_dump" + ) + init_svn_repo_from_dump.return_value = ("", "") + mock_client = mocker.MagicMock() + mocker.patch.object(client, "Client", mock_client) + + class Info: + repos_root_url = repo_redirect_url + url = repo_redirect_url + + mock_client().info.return_value = {"repo": Info()} + + # init remote dump loader and mock some methods + loader = SvnLoaderFromRemoteDump(swh_storage, repo_url, temp_directory=tmp_path) + loader.dump_svn_revisions = mocker.MagicMock() + loader.start_from = mocker.MagicMock(return_value=(0, 0)) + + # prepare loading + loader.prepare() + + # check redirection URL has been used to dump repository + assert loader.dump_svn_revisions.call_args_list[0][0][0] == repo_redirect_url