diff --git a/swh/loader/svn/loader.py b/swh/loader/svn/loader.py index d5342e0..ae6ff91 100644 --- a/swh/loader/svn/loader.py +++ b/swh/loader/svn/loader.py @@ -1,852 +1,857 @@ # 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 """Loader in charge of injecting either new or existing svn mirrors to swh-storage. """ from datetime import datetime import os import pty import re import shutil from subprocess import PIPE, 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 ) -> from_disk.Directory: """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 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, dir: from_disk.Directory ) -> 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 = self.swh_revision_hash_tree_at_svn_revision(rev) checked_dir_id = checked_dir.hash if checked_dir_id != dir_id: # do not bother checking tree differences if root directory id of reconstructed # repository filesystem does not match the id of the one from the last loaded # revision (can happen when called from post_load and tree differences were checked # before the last revision to load) if self.debug and dir_id == dir.hash: for obj in checked_dir.iter_tree(): path = obj.data["path"].replace(checked_dir.data["path"], b"")[1:] if not path: # ignore root directory continue if path not in dir: self.log.debug( "%s with path %s is missing in reconstructed repository filesystem", obj.object_type, # type: ignore path, ) elif dir[path].hash != checked_dir[path].hash: self.log.debug( "%s with path %s has different hash in reconstructed repository filesystem", # noqa obj.object_type, # type: ignore path, ) err = ( "Hash tree computation divergence detected at revision %s " "(%s != %s), stopping!" % ( rev, 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, root_directory) 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, + debug=self.debug, ) 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, self.svnrepo.swhreplay.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 ) -> Tuple[str, int]: """Generate a compressed subversion dump file using the svnrdump tool and gzip. 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 and the max dumped revision number (-1 if all revisions were dumped) """ # Build the svnrdump command line svnrdump_cmd = ["svnrdump", "dump", svn_url] assert self.svnrepo is not None if self.svnrepo.username: svnrdump_cmd += [ "--username", self.svnrepo.username, "--password", self.svnrepo.password, ] # 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.gz" % (dump_temp_dir, dump_name) stderr_lines = [] self.log.debug("Executing %s", " ".join(svnrdump_cmd)) with open(dump_path, "wb") as dump_file: gzip = Popen(["gzip"], stdin=PIPE, stdout=dump_file) stderr_r, stderr_w = pty.openpty() svnrdump = Popen(svnrdump_cmd, stdout=gzip.stdin, 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) # denote end of read file gzip.stdin.close() gzip.wait() if svnrdump.returncode == 0: return dump_path, -1 # 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, ) self.truncated_dump = True return dump_path, last_dumped_rev 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.origin.url) self.svnrepo = self.svn_repo( - self.origin.url, self.origin.url, self.temp_dir, self.max_content_size + self.origin.url, + self.origin.url, + self.temp_dir, + self.max_content_size, + debug=self.debug, ) # 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 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, max_rev = 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, gzip=True, max_rev=max_rev, ) 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/replay.py b/swh/loader/svn/replay.py index febd5e8..fd11be3 100644 --- a/swh/loader/svn/replay.py +++ b/swh/loader/svn/replay.py @@ -1,1088 +1,1129 @@ # 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 """Remote Access client to svn server. """ from __future__ import annotations import codecs from collections import defaultdict from dataclasses import dataclass, field from distutils.dir_util import copy_tree from itertools import chain import logging import os import shutil import tempfile from typing import ( TYPE_CHECKING, Any, BinaryIO, Callable, Dict, List, Optional, Set, Tuple, Union, cast, ) import click from subvertpy import SubversionException, delta, properties from subvertpy.ra import Auth, RemoteAccess, get_username_provider from swh.model import from_disk, hashutil from swh.model.from_disk import DiskBackedContent from swh.model.model import Content, Directory, SkippedContent if TYPE_CHECKING: from swh.loader.svn.svn import SvnRepo from swh.loader.svn.utils import ( is_recursive_external, parse_external_definition, svn_urljoin, ) _eol_style = {"native": b"\n", "CRLF": b"\r\n", "LF": b"\n", "CR": b"\r"} logger = logging.getLogger(__name__) def _normalize_line_endings(lines: bytes, eol_style: str = "native") -> bytes: r"""Normalize line endings to unix (\\n), windows (\\r\\n) or mac (\\r). Args: lines: The lines to normalize eol_style: The line ending format as defined for svn:eol-style property. Acceptable values are 'native', 'CRLF', 'LF' and 'CR' Returns: Lines with endings normalized """ if eol_style in _eol_style: lines = lines.replace(_eol_style["CRLF"], _eol_style["LF"]).replace( _eol_style["CR"], _eol_style["LF"] ) if _eol_style[eol_style] != _eol_style["LF"]: lines = lines.replace(_eol_style["LF"], _eol_style[eol_style]) return lines def apply_txdelta_handler( sbuf: bytes, target_stream: BinaryIO ) -> Callable[[Any, bytes, BinaryIO], None]: """Return a function that can be called repeatedly with txdelta windows. When done, closes the target_stream. Adapted from subvertpy.delta.apply_txdelta_handler to close the stream when done. Args: sbuf: Source buffer target_stream: Target stream to write to. Returns: Function to be called to apply txdelta windows """ def apply_window( window: Any, sbuf: bytes = sbuf, target_stream: BinaryIO = target_stream ): if window is None: target_stream.close() return # Last call patch = delta.apply_txdelta_window(sbuf, window) target_stream.write(patch) return apply_window def read_svn_link(data: bytes) -> Tuple[bytes, bytes]: """Read the svn link's content. Args: data: svn link's raw content Returns: The tuple of (filetype, destination path) """ split_byte = b" " first_line = data.split(b"\n")[0] filetype, *src = first_line.split(split_byte) target = split_byte.join(src) return filetype, target def is_file_an_svnlink_p(fullpath: bytes) -> Tuple[bool, bytes]: """Determine if a filepath is an svnlink or something else. Args: fullpath: Full path to the potential symlink to check Returns: Tuple containing a boolean value to determine if it's indeed a symlink (as per svn) and the link target. """ if os.path.islink(fullpath): return False, b"" with open(fullpath, "rb") as f: filetype, src = read_svn_link(f.read()) return filetype == b"link", src def _ra_codecs_error_handler(e: UnicodeError) -> Tuple[Union[str, bytes], int]: """Subvertpy may fail to decode to utf-8 the user svn properties. As they are not used by the loader, return an empty string instead of the decoded content. Args: e: exception raised during the svn properties decoding. """ return "", cast(UnicodeDecodeError, e).end DEFAULT_FLAG = 0 EXEC_FLAG = 1 NOEXEC_FLAG = 2 SVN_PROPERTY_EOL = "svn:eol-style" @dataclass class FileState: """Persists some file states (eg. end of lines style) across revisions while replaying them.""" eol_style: Optional[str] = None """EOL state check mess""" svn_special_path_non_link_data: Optional[bytes] = None """keep track of non link file content with svn:special property set""" # default value: 0, 1: set the flag, 2: remove the exec flag executable: int = DEFAULT_FLAG """keep track if file is executable when setting svn:executable property""" link: bool = False """keep track if file is a svn link when setting svn:special property""" class FileEditor: """File Editor in charge of updating file on disk and memory objects.""" __slots__ = [ "directory", "path", "fullpath", "executable", "link", "state", "svnrepo", "editor", ] def __init__( self, directory: from_disk.Directory, rootpath: bytes, path: bytes, state: FileState, svnrepo: SvnRepo, ): self.directory = directory self.path = path self.fullpath = os.path.join(rootpath, path) self.state = state self.svnrepo = svnrepo self.editor = svnrepo.swhreplay.editor def change_prop(self, key: str, value: str) -> None: + if self.editor.debug: + logger.debug( + "Setting property %s to value %s on path %s", key, value, self.path + ) if key == properties.PROP_EXECUTABLE: if value is None: # bit flip off self.state.executable = NOEXEC_FLAG else: self.state.executable = EXEC_FLAG elif key == properties.PROP_SPECIAL: # Possibly a symbolic link. We cannot check further at # that moment though, patch(s) not being applied yet self.state.link = value is not None elif key == SVN_PROPERTY_EOL: # backup end of line style for file self.state.eol_style = value def __make_symlink(self, src: bytes) -> None: """Convert the svnlink to a symlink on disk. This function expects self.fullpath to be a svn link. Args: src: Path to the link's source Return: tuple: The svnlink's data tuple: - type (should be only 'link') - """ os.remove(self.fullpath) os.symlink(src=src, dst=self.fullpath) def __make_svnlink(self) -> bytes: """Convert the symlink to a svnlink on disk. Return: The symlink's svnlink data (``b'type '``) """ # we replace the symlink by a svnlink # to be able to patch the file on future commits src = os.readlink(self.fullpath) os.remove(self.fullpath) sbuf = b"link " + src with open(self.fullpath, "wb") as f: f.write(sbuf) return sbuf def apply_textdelta(self, base_checksum) -> Callable[[Any, bytes, BinaryIO], None]: + if self.editor.debug: + logger.debug("Applying textdelta to file %s", self.path) # if the filepath matches an external, do not apply local patch if self.path in self.editor.external_paths: return lambda *args: None if os.path.lexists(self.fullpath): if os.path.islink(self.fullpath): # svn does not deal with symlink so we transform into # real svn symlink for potential patching in later # commits sbuf = self.__make_svnlink() self.state.link = True else: with open(self.fullpath, "rb") as f: sbuf = f.read() else: sbuf = b"" t = open(self.fullpath, "wb") return apply_txdelta_handler(sbuf, target_stream=t) def close(self) -> None: """When done with the file, this is called. So the file exists and is updated, we can: - adapt accordingly its execution flag if any - compute the objects' checksums - replace the svnlink with a real symlink (for disk computation purposes) """ + if self.editor.debug: + logger.debug("Closing file %s", self.path) if self.state.link: # can only check now that the link is a real one # since patch has been applied is_link, src = is_file_an_svnlink_p(self.fullpath) if is_link: self.__make_symlink(src) elif not os.path.isdir(self.fullpath): # not a real link ... # when a file with the svn:special property set is not a svn link, # the svn export operation might extract a truncated version of it # if it is a binary file, so ensure to produce the same file as the # export operation. with open(self.fullpath, "rb") as f: content = f.read() self.svnrepo.export( os.path.join(self.svnrepo.remote_url, os.fsdecode(self.path)), to=self.fullpath, peg_rev=self.editor.revnum, ignore_keywords=True, overwrite=True, ) with open(self.fullpath, "rb") as f: exported_data = f.read() if exported_data != content: # keep track of original file content in order to restore # it if the svn:special property gets unset in another revision self.state.svn_special_path_non_link_data = content elif os.path.islink(self.fullpath): # path was a symbolic link in previous revision but got the property # svn:special unset in current one, revert its content to svn link format self.__make_svnlink() elif self.state.svn_special_path_non_link_data is not None: # path was a non link file with the svn:special property previously set # and got truncated on export, restore its original content with open(self.fullpath, "wb") as f: f.write(self.state.svn_special_path_non_link_data) self.state.svn_special_path_non_link_data = None is_link = os.path.islink(self.fullpath) if not is_link: # if a link, do nothing regarding flag if self.state.executable == EXEC_FLAG: os.chmod(self.fullpath, 0o755) elif self.state.executable == NOEXEC_FLAG: os.chmod(self.fullpath, 0o644) # And now compute file's checksums if self.state.eol_style and not is_link: # ensure to normalize line endings as defined by svn:eol-style # property to get the same file checksum as after an export # or checkout operation with subversion with open(self.fullpath, "rb") as f: data = f.read() data = _normalize_line_endings(data, self.state.eol_style) mode = os.lstat(self.fullpath).st_mode self.directory[self.path] = from_disk.Content.from_bytes( mode=mode, data=data ) else: self.directory[self.path] = from_disk.Content.from_file(path=self.fullpath) ExternalDefinition = Tuple[str, Optional[int], bool] @dataclass class DirState: """Persists some directory states (eg. externals) across revisions while replaying them.""" externals: Dict[str, List[ExternalDefinition]] = field(default_factory=dict) """Map a path in the directory to a list of (external_url, revision, relative_url) targeting it""" externals_paths: Set[bytes] = field(default_factory=set) """Keep track of all external paths reachable from the directory""" class DirEditor: """Directory Editor in charge of updating directory hashes computation. This implementation includes empty folder in the hash computation. """ __slots__ = [ "directory", "rootpath", "path", "file_states", "dir_states", "svnrepo", "editor", "externals", ] def __init__( self, directory: from_disk.Directory, rootpath: bytes, path: bytes, file_states: Dict[bytes, FileState], dir_states: Dict[bytes, DirState], svnrepo: SvnRepo, ): self.directory = directory self.rootpath = rootpath self.path = path # build directory on init os.makedirs(rootpath, exist_ok=True) self.file_states = file_states self.dir_states = dir_states self.svnrepo = svnrepo self.editor = svnrepo.swhreplay.editor self.externals: Dict[str, List[ExternalDefinition]] = {} def remove_child(self, path: bytes) -> None: """Remove a path from the current objects. The path can be resolved as link, file or directory. This function takes also care of removing the link between the child and the parent. Args: path: to remove from the current objects. """ if path in self.directory: entry_removed = self.directory[path] del self.directory[path] fpath = os.path.join(self.rootpath, path) if isinstance(entry_removed, from_disk.Directory): shutil.rmtree(fpath) else: os.remove(fpath) # when deleting a directory ensure to remove any svn property for the # file it contains as they can be added again later in another revision # without the same property set fullpath = os.path.join(self.rootpath, path) for state_path in list(self.file_states): if state_path.startswith(fullpath + b"/"): del self.file_states[state_path] def open_directory(self, path: str, *args) -> DirEditor: """Updating existing directory.""" + if self.editor.debug: + logger.debug("Opening directory %s", path) return DirEditor( self.directory, rootpath=self.rootpath, path=os.fsencode(path), file_states=self.file_states, dir_states=self.dir_states, svnrepo=self.svnrepo, ) def add_directory( self, path: str, copyfrom_path: Optional[str] = None, copyfrom_rev: int = -1 ) -> DirEditor: """Adding a new directory.""" + if self.editor.debug: + logger.debug( + "Adding directory %s, copyfrom_path = %s, copyfrom_rev = %s", + path, + copyfrom_path, + copyfrom_rev, + ) + path_bytes = os.fsencode(path) fullpath = os.path.join(self.rootpath, path_bytes) os.makedirs(fullpath, exist_ok=True) if copyfrom_rev == -1: if path_bytes and path_bytes not in self.directory: self.dir_states[path_bytes] = DirState() self.directory[path_bytes] = from_disk.Directory() else: url = svn_urljoin(self.svnrepo.remote_url, copyfrom_path) self.remove_child(path_bytes) self.svnrepo.export( url, to=fullpath, peg_rev=copyfrom_rev, ignore_keywords=True, overwrite=True, ignore_externals=True, ) self.directory[path_bytes] = from_disk.Directory.from_disk(path=fullpath) return DirEditor( self.directory, self.rootpath, path_bytes, self.file_states, self.dir_states, svnrepo=self.svnrepo, ) def open_file(self, path: str, *args) -> FileEditor: """Updating existing file.""" + if self.editor.debug: + logger.debug("Opening file %s", path) + path_bytes = os.fsencode(path) self.directory[path_bytes] = from_disk.Content() fullpath = os.path.join(self.rootpath, path_bytes) return FileEditor( self.directory, rootpath=self.rootpath, path=path_bytes, state=self.file_states[fullpath], svnrepo=self.svnrepo, ) def add_file( self, path: str, copyfrom_path: Optional[str] = None, copyfrom_rev: int = -1 ) -> FileEditor: """Creating a new file.""" + if self.editor.debug: + logger.debug( + "Adding file %s, copyfrom_path = %s, copyfrom_rev = %s", + path, + copyfrom_path, + copyfrom_rev, + ) + path_bytes = os.fsencode(path) fullpath = os.path.join(self.rootpath, path_bytes) self.file_states[fullpath] = FileState() if copyfrom_rev == -1: self.directory[path_bytes] = from_disk.Content() else: url = svn_urljoin(self.svnrepo.remote_url, copyfrom_path) self.remove_child(path_bytes) self.svnrepo.export( url, to=fullpath, peg_rev=copyfrom_rev, ignore_keywords=True, overwrite=True, ) self.directory[path_bytes] = from_disk.Content.from_file(path=fullpath) return FileEditor( self.directory, self.rootpath, path_bytes, state=self.file_states[fullpath], svnrepo=self.svnrepo, ) def change_prop(self, key: str, value: str) -> None: """Change property callback on directory.""" if key == properties.PROP_EXTERNALS: logger.debug( "Setting '%s' property with value '%s' on path %s", key, value, self.path, ) self.externals = defaultdict(list) if value is not None: try: # externals are set on that directory path, parse and store them # for later processing in the close method for external in value.split("\n"): external = external.rstrip("\r") # skip empty line or comment if not external or external.startswith("#"): continue ( path, external_url, revision, relative_url, ) = parse_external_definition( external, os.fsdecode(self.path), self.svnrepo.origin_url ) self.externals[path].append( (external_url, revision, relative_url) ) except ValueError: logger.debug( "Failed to parse external: %s\n" "Externals defined on path %s will not be processed", external, self.path, ) # as the official subversion client, do not process externals in case # of parsing error self.externals = {} if not self.externals: # externals might have been unset on that directory path, # remove associated paths from the reconstructed filesystem externals = self.dir_states[self.path].externals for path in externals.keys(): self.remove_external_path(os.fsencode(path)) self.dir_states[self.path].externals = {} def delete_entry(self, path: str, revision: int) -> None: """Remove a path.""" + if self.editor.debug: + logger.debug("Deleting directory entry %s", path) + path_bytes = os.fsencode(path) fullpath = os.path.join(self.rootpath, path_bytes) if os.path.isdir(fullpath): # remove all external paths associated to the removed directory # (we cannot simply remove a root external directory as externals # paths associated to ancestor directories can overlap) for external_path in self.dir_states[path_bytes].externals_paths: self.remove_external_path( external_path, root_path=path_bytes, remove_subpaths=False, force=True, ) if path_bytes not in self.editor.external_paths: self.file_states.pop(fullpath, None) self.remove_child(path_bytes) elif os.path.isdir(fullpath): # versioned and external paths can overlap so we need to iterate on # all subpaths to check which ones to remove for root, dirs, files in os.walk(fullpath): for p in chain(dirs, files): full_repo_path = os.path.join(root, p) repo_path = full_repo_path.replace(self.rootpath + b"/", b"") if repo_path not in self.editor.external_paths: self.file_states.pop(full_repo_path, None) self.remove_child(repo_path) def close(self): """Function called when we finish processing a repository. SVN external definitions are processed by it. """ + if self.editor.debug: + logger.debug("Closing directory %s", self.path) prev_externals = self.dir_states[self.path].externals if self.externals: # externals definition list might have changed in the current replayed # revision, we need to determine if some were removed and delete the # associated paths externals = self.externals prev_externals_set = { (path, url, rev) for path in prev_externals.keys() for (url, rev, _) in prev_externals[path] } externals_set = { (path, url, rev) for path in externals.keys() for (url, rev, _) in externals[path] } old_externals = prev_externals_set - externals_set for path, _, _ in old_externals: self.remove_external_path(os.fsencode(path)) else: # some external paths might have been removed in the current replayed # revision by a delete operation on an overlapping versioned path so we # need to restore them externals = prev_externals # For each external, try to export it in reconstructed filesystem for path, externals_def in externals.items(): for i, external in enumerate(externals_def): external_url, revision, relative_url = external self.process_external( path, external_url, revision, relative_url, remove_target_path=i == 0, ) # backup externals in directory state if self.externals: self.dir_states[self.path].externals = self.externals # do operations below only when closing the root directory if self.path == b"": self.svnrepo.has_relative_externals = any( relative_url for (_, relative_url) in self.editor.valid_externals.values() ) self.svnrepo.has_recursive_externals = any( is_recursive_external( self.svnrepo.origin_url, os.fsdecode(path), external_path, external_url, ) for path, dir_state in self.dir_states.items() for external_path in dir_state.externals.keys() for (external_url, _, _) in dir_state.externals[external_path] ) if self.svnrepo.has_recursive_externals: # If the repository has recursive externals, we stop processing # externals and remove those already exported, # We will then ignore externals when exporting the revision to # check for divergence with the reconstructed filesystem. for external_path in list(self.editor.external_paths): self.remove_external_path(external_path, force=True) def process_external( self, path: str, external_url: str, revision: Optional[int], relative_url: bool, remove_target_path: bool = True, ) -> None: external = (external_url, revision, relative_url) dest_path = os.fsencode(path) dest_fullpath = os.path.join(self.path, dest_path) prev_externals = self.dir_states[self.path].externals if ( path in prev_externals and external in prev_externals[path] and dest_fullpath in self.directory ): # external already exported, nothing to do return if is_recursive_external( self.svnrepo.origin_url, os.fsdecode(self.path), path, external_url ): # recursive external, skip it return logger.debug( "Exporting external %s%s to path %s", external_url, f"@{revision}" if revision else "", dest_fullpath, ) if external not in self.editor.externals_cache: try: # try to export external in a temporary path, destination path could # be versioned and must be overridden only if the external URL is # still valid temp_dir = os.fsencode( tempfile.mkdtemp(dir=self.editor.externals_cache_dir) ) temp_path = os.path.join(temp_dir, dest_path) os.makedirs(b"/".join(temp_path.split(b"/")[:-1]), exist_ok=True) if external_url not in self.editor.dead_externals: url = external_url.rstrip("/") origin_url = self.svnrepo.origin_url.rstrip("/") if ( url.startswith(origin_url + "/") and not self.svnrepo.has_relative_externals ): url = url.replace(origin_url, self.svnrepo.remote_url) self.svnrepo.export( url, to=temp_path, peg_rev=revision, ignore_keywords=True, ) self.editor.externals_cache[external] = temp_path except SubversionException as se: # external no longer available (404) logger.debug(se) self.editor.dead_externals.add(external_url) else: temp_path = self.editor.externals_cache[external] # subversion export will always create the subdirectories of the external # path regardless the validity of the remote URL dest_path_split = dest_path.split(b"/") current_path = self.path self.add_directory(os.fsdecode(current_path)) for subpath in dest_path_split[:-1]: current_path = os.path.join(current_path, subpath) self.add_directory(os.fsdecode(current_path)) if os.path.exists(temp_path): # external successfully exported if remove_target_path: # remove previous path in from_disk model self.remove_external_path(dest_path, remove_subpaths=False) # mark external as valid self.editor.valid_externals[dest_fullpath] = ( external_url, relative_url, ) # copy exported path to reconstructed filesystem fullpath = os.path.join(self.rootpath, dest_fullpath) if os.path.isfile(temp_path): if os.path.islink(fullpath): # remove destination file if it is a link os.remove(fullpath) shutil.copy(os.fsdecode(temp_path), os.fsdecode(fullpath)) self.directory[dest_fullpath] = from_disk.Content.from_file( path=fullpath ) else: self.add_directory(os.fsdecode(dest_fullpath)) # copy_tree needs sub-directories to exist in destination for root, dirs, files in os.walk(temp_path): for dir in dirs: temp_dir_fullpath = os.path.join(root, dir) if os.path.islink(temp_dir_fullpath): # do not create folder if it's a link or copy_tree will fail continue subdir = temp_dir_fullpath.replace(temp_path + b"/", b"") self.add_directory( os.fsdecode(os.path.join(dest_fullpath, subdir)) ) copy_tree( os.fsdecode(temp_path), os.fsdecode(fullpath), preserve_symlinks=True, ) # TODO: replace code above by the line below once we use Python >= 3.8 in production # noqa # shutil.copytree(temp_path, fullpath, symlinks=True, dirs_exist_ok=True) # noqa self.directory[dest_fullpath] = from_disk.Directory.from_disk( path=fullpath ) # update set of external paths reachable from the directory external_paths = set() dest_path_part = dest_path.split(b"/") for i in range(1, len(dest_path_part) + 1): external_paths.add(b"/".join(dest_path_part[:i])) for root, dirs, files in os.walk(temp_path): external_paths.update( [ os.path.join( dest_path, os.path.join(root, p).replace(temp_path, b"").strip(b"/"), ) for p in chain(dirs, files) ] ) self.dir_states[self.path].externals_paths.update(external_paths) for external_path in external_paths: self.editor.external_paths[os.path.join(self.path, external_path)] += 1 # ensure hash update for the directory with externals set self.directory[self.path].update_hash(force=True) def remove_external_path( self, external_path: bytes, remove_subpaths: bool = True, force: bool = False, root_path: Optional[bytes] = None, ) -> None: """Remove a previously exported SVN external path from the reconstructed filesystem. """ path = root_path if root_path else self.path fullpath = os.path.join(path, external_path) # decrement number of references for external path when we really remove it # (when remove_subpaths is False, we just cleanup the external path before # copying exported paths in it) if force or (fullpath in self.editor.external_paths and remove_subpaths): self.editor.external_paths[fullpath] -= 1 if ( fullpath in self.editor.external_paths and self.editor.external_paths[fullpath] == 0 ): self.remove_child(fullpath) self.editor.external_paths.pop(fullpath, None) self.editor.valid_externals.pop(fullpath, None) for path in list(self.editor.external_paths): if path.startswith(fullpath + b"/"): self.editor.external_paths[path] -= 1 if self.editor.external_paths[path] == 0: self.editor.external_paths.pop(path) if remove_subpaths: subpath_split = fullpath.split(b"/")[:-1] for i in reversed(range(1, len(subpath_split) + 1)): # delete external sub-directory only if it is not versioned subpath = b"/".join(subpath_split[0:i]) try: self.svnrepo.client.info( svn_urljoin(self.svnrepo.remote_url, os.fsdecode(subpath)), peg_revision=self.editor.revnum, revision=self.editor.revnum, ) except SubversionException: self.remove_child(subpath) else: break try: # externals can overlap with versioned files so we must restore # them after removing the path above dest_path = os.path.join(self.rootpath, fullpath) self.svnrepo.client.export( svn_urljoin(self.svnrepo.remote_url, os.fsdecode(fullpath)), to=dest_path, peg_rev=self.editor.revnum, ignore_keywords=True, ) if os.path.isfile(dest_path) or os.path.islink(dest_path): self.directory[fullpath] = from_disk.Content.from_file(path=dest_path) else: self.directory[fullpath] = from_disk.Directory.from_disk(path=dest_path) except SubversionException: pass class Editor: """Editor in charge of replaying svn events and computing objects along. This implementation accounts for empty folder during hash computations. """ def __init__( self, rootpath: bytes, directory: from_disk.Directory, svnrepo: SvnRepo, temp_dir: str, + debug: bool = False, ): self.rootpath = rootpath self.directory = directory self.file_states: Dict[bytes, FileState] = defaultdict(FileState) self.dir_states: Dict[bytes, DirState] = defaultdict(DirState) self.external_paths: Dict[bytes, int] = defaultdict(int) self.valid_externals: Dict[bytes, Tuple[str, bool]] = {} self.dead_externals: Set[str] = set() self.externals_cache_dir = tempfile.mkdtemp(dir=temp_dir) self.externals_cache: Dict[ExternalDefinition, bytes] = {} self.svnrepo = svnrepo self.revnum = None + self.debug = debug def set_target_revision(self, revnum) -> None: self.revnum = revnum def abort(self) -> None: pass def close(self) -> None: pass def open_root(self, base_revnum: int) -> DirEditor: return DirEditor( self.directory, rootpath=self.rootpath, path=b"", file_states=self.file_states, dir_states=self.dir_states, svnrepo=self.svnrepo, ) class Replay: """Replay class.""" def __init__( self, conn: RemoteAccess, rootpath: bytes, svnrepo: SvnRepo, temp_dir: str, directory: Optional[from_disk.Directory] = None, + debug: bool = False, ): self.conn = conn self.rootpath = rootpath if directory is None: directory = from_disk.Directory() self.directory = directory self.editor = Editor( - rootpath=rootpath, directory=directory, svnrepo=svnrepo, temp_dir=temp_dir + rootpath=rootpath, + directory=directory, + svnrepo=svnrepo, + temp_dir=temp_dir, + debug=debug, ) def replay(self, rev: int, low_water_mark: int) -> from_disk.Directory: """Replay svn actions between rev and rev+1. This method updates in place the self.editor.directory, as well as the filesystem. Returns: The updated root directory """ codecs.register_error("strict", _ra_codecs_error_handler) self.conn.replay(rev, low_water_mark, self.editor) codecs.register_error("strict", codecs.strict_errors) return self.editor.directory def compute_objects( self, rev: int, low_water_mark: int ) -> Tuple[List[Content], List[SkippedContent], List[Directory]]: """Compute objects added or modified at revisions rev. Expects the state to be at previous revision's objects. Args: rev: The revision to start the replay from. Returns: The updated objects between rev and rev+1. Beware that this mutates the filesystem at rootpath accordingly. """ self.replay(rev, low_water_mark) contents: List[Content] = [] skipped_contents: List[SkippedContent] = [] directories: List[Directory] = [] for obj_node in self.directory.collect(): obj = obj_node.to_model() # type: ignore obj_type = obj.object_type if obj_type in (Content.object_type, DiskBackedContent.object_type): contents.append(obj.with_data()) elif obj_type == SkippedContent.object_type: skipped_contents.append(obj) elif obj_type == Directory.object_type: directories.append(obj) else: assert False, obj_type return contents, skipped_contents, directories @click.command() @click.option("--local-url", default="/tmp", help="local svn working copy") @click.option( "--svn-url", default="file:///home/storage/svn/repos/pkg-fox", help="svn repository's url.", ) @click.option( "--revision-start", default=1, type=click.INT, help="svn repository's starting revision.", ) @click.option( "--revision-end", default=-1, type=click.INT, help="svn repository's ending revision.", ) @click.option( "--debug/--nodebug", default=True, help="Indicates if the server should run in debug mode.", ) @click.option( "--cleanup/--nocleanup", default=True, help="Indicates whether to cleanup disk when done or not.", ) def main(local_url, svn_url, revision_start, revision_end, debug, cleanup): """Script to present how to use Replay class.""" conn = RemoteAccess(svn_url.encode("utf-8"), auth=Auth([get_username_provider()])) os.makedirs(local_url, exist_ok=True) rootpath = tempfile.mkdtemp( prefix=local_url, suffix="-" + os.path.basename(svn_url) ) rootpath = os.fsencode(rootpath) # Do not go beyond the repository's latest revision revision_end_max = conn.get_latest_revnum() if revision_end == -1: revision_end = revision_end_max revision_end = min(revision_end, revision_end_max) try: replay = Replay(conn, rootpath) for rev in range(revision_start, revision_end + 1): contents, skipped_contents, directories = replay.compute_objects(rev) print( "r%s %s (%s new contents, %s new directories)" % ( rev, hashutil.hash_to_hex(replay.directory.hash), len(contents) + len(skipped_contents), len(directories), ) ) if debug: print("%s" % rootpath.decode("utf-8")) finally: if cleanup: if os.path.exists(rootpath): shutil.rmtree(rootpath) if __name__ == "__main__": main() diff --git a/swh/loader/svn/svn.py b/swh/loader/svn/svn.py index 1593ef7..f88d523 100644 --- a/swh/loader/svn/svn.py +++ b/swh/loader/svn/svn.py @@ -1,611 +1,613 @@ # 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 urllib.parse import urlparse, urlunparse from subvertpy import SubversionException, client, properties, wc from subvertpy.ra import ( Auth, RemoteAccess, get_simple_prompt_provider, 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__) def quote_svn_url(url: str) -> str: return url.replace(" ", "%20").replace("#", "%23") 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, + debug: bool = False, ): self.origin_url = origin_url self.from_dump = from_dump # default auth provider for anonymous access auth_providers = [get_username_provider()] # check if basic auth is required parsed_origin_url = urlparse(origin_url) self.username = parsed_origin_url.username or "" self.password = parsed_origin_url.password or "" if self.username: # add basic auth provider for username/password auth_providers.append( get_simple_prompt_provider( lambda realm, uname, may_save: ( self.username, self.password, False, ), 0, ) ) # we need to remove the authentication part in the origin URL to avoid # errors when calling subversion API through subvertpy self.origin_url = urlunparse( parsed_origin_url._replace( netloc=parsed_origin_url.netloc.split("@", 1)[1] ) ) if origin_url == remote_url: remote_url = self.origin_url self.remote_url = remote_url.rstrip("/") auth = Auth(auth_providers) # 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) 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, + debug=debug, ) 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(self.origin_url).repos_root_url self.root_directory = self.origin_url.rstrip("/").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, "changed_paths": changed_paths, } 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=True, ): yield self.__to_entry(log_entry) @svn_retry() def commit_info(self, revision: int) -> Optional[Dict]: """Return commit information. Args: revision: svn revision to return commit info Returns: A dictionary filled with commit info, see :meth:`swh.loader.svn.svn.logs` for details about its content. """ return next(self.logs(revision, revision), None) @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(quote_svn_url(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), quote_svn_url(url), f"@{peg_rev}" if peg_rev else "", to, ) return self.client.export( quote_svn_url(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), quote_svn_url(url), f"@{peg_rev}" if peg_rev else "", path, ) return self.client.checkout( quote_svn_url(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"] copyfrom_revs = ( [ copyfrom_rev for (_, _, copyfrom_rev, _) in commit["changed_paths"].values() if copyfrom_rev != -1 ] if commit["changed_paths"] else None ) low_water_mark = rev + 1 if copyfrom_revs: # when files or directories in the revision to replay have been copied from # ancestor revisions, we need to adjust the low water mark revision used by # svn replay API to handle the copies in our commit editor and to ensure # replace operations after copy will be replayed low_water_mark = min(copyfrom_revs) objects = self.swhreplay.compute_objects(rev, low_water_mark) 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 = self.commit_info(revision) # 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)