diff --git a/PKG-INFO b/PKG-INFO index a68e999..798ab38 100644 --- a/PKG-INFO +++ b/PKG-INFO @@ -1,61 +1,61 @@ Metadata-Version: 2.1 Name: swh.loader.svn -Version: 0.10.2 +Version: 1.0.0 Summary: Software Heritage Loader SVN Home-page: https://forge.softwareheritage.org/diffusion/DLDSVN Author: Software Heritage developers Author-email: swh-devel@inria.fr License: UNKNOWN Project-URL: Bug Reports, https://forge.softwareheritage.org/maniphest Project-URL: Funding, https://www.softwareheritage.org/donate Project-URL: Source, https://forge.softwareheritage.org/source/swh-loader-svn Project-URL: Documentation, https://docs.softwareheritage.org/devel/swh-loader-svn/ Platform: UNKNOWN Classifier: Programming Language :: Python :: 3 Classifier: Intended Audience :: Developers Classifier: License :: OSI Approved :: GNU General Public License v3 (GPLv3) Classifier: Operating System :: OS Independent Classifier: Development Status :: 5 - Production/Stable Requires-Python: >=3.7 Description-Content-Type: text/markdown Provides-Extra: testing License-File: LICENSE License-File: AUTHORS swh-loader-svn ============== The Software Heritage SVN Loader is a tool and a library to walk a remote svn repository and inject into the SWH dataset all contained files that weren't known before. The main entry points are - :class:`swh.loader.svn.loader.SvnLoader` for the main svn loader which ingests content out of a remote svn repository - :class:`swh.loader.svn.loader.SvnLoaderFromDumpArchive` which mounts a repository out of a svn dump prior to ingest it. - :class:`swh.loader.svn.loader.SvnLoaderFromRemoteDump` which mounts a repository with svnrdump prior to ingest its content. # CLI run With the configuration: /tmp/loader_svn.yml: ``` storage: cls: remote args: url: http://localhost:5002/ ``` Run: ``` swh loader --config-file /tmp/loader_svn.yml \ run svn ``` diff --git a/requirements-swh.txt b/requirements-swh.txt index 2abb671..33a001b 100644 --- a/requirements-swh.txt +++ b/requirements-swh.txt @@ -1,4 +1,4 @@ swh.storage >= 0.11.3 -swh.model >= 0.4.0 +swh.model >= 4.3.0 swh.scheduler >= 0.0.39 swh.loader.core >= 0.18 diff --git a/requirements.txt b/requirements.txt index 5730ae2..f03579d 100644 --- a/requirements.txt +++ b/requirements.txt @@ -1,8 +1,7 @@ # Add here external Python modules dependencies, one per line. Module names # should match https://pypi.python.org/pypi names. For the full spec or # dependency lines, see https://pip.readthedocs.org/en/1.1/requirements.html click iso8601 -python-dateutil subvertpy >= 0.9.4 typing-extensions diff --git a/swh.loader.svn.egg-info/PKG-INFO b/swh.loader.svn.egg-info/PKG-INFO index a68e999..798ab38 100644 --- a/swh.loader.svn.egg-info/PKG-INFO +++ b/swh.loader.svn.egg-info/PKG-INFO @@ -1,61 +1,61 @@ Metadata-Version: 2.1 Name: swh.loader.svn -Version: 0.10.2 +Version: 1.0.0 Summary: Software Heritage Loader SVN Home-page: https://forge.softwareheritage.org/diffusion/DLDSVN Author: Software Heritage developers Author-email: swh-devel@inria.fr License: UNKNOWN Project-URL: Bug Reports, https://forge.softwareheritage.org/maniphest Project-URL: Funding, https://www.softwareheritage.org/donate Project-URL: Source, https://forge.softwareheritage.org/source/swh-loader-svn Project-URL: Documentation, https://docs.softwareheritage.org/devel/swh-loader-svn/ Platform: UNKNOWN Classifier: Programming Language :: Python :: 3 Classifier: Intended Audience :: Developers Classifier: License :: OSI Approved :: GNU General Public License v3 (GPLv3) Classifier: Operating System :: OS Independent Classifier: Development Status :: 5 - Production/Stable Requires-Python: >=3.7 Description-Content-Type: text/markdown Provides-Extra: testing License-File: LICENSE License-File: AUTHORS swh-loader-svn ============== The Software Heritage SVN Loader is a tool and a library to walk a remote svn repository and inject into the SWH dataset all contained files that weren't known before. The main entry points are - :class:`swh.loader.svn.loader.SvnLoader` for the main svn loader which ingests content out of a remote svn repository - :class:`swh.loader.svn.loader.SvnLoaderFromDumpArchive` which mounts a repository out of a svn dump prior to ingest it. - :class:`swh.loader.svn.loader.SvnLoaderFromRemoteDump` which mounts a repository with svnrdump prior to ingest its content. # CLI run With the configuration: /tmp/loader_svn.yml: ``` storage: cls: remote args: url: http://localhost:5002/ ``` Run: ``` swh loader --config-file /tmp/loader_svn.yml \ run svn ``` diff --git a/swh.loader.svn.egg-info/SOURCES.txt b/swh.loader.svn.egg-info/SOURCES.txt index 3239751..52adf48 100644 --- a/swh.loader.svn.egg-info/SOURCES.txt +++ b/swh.loader.svn.egg-info/SOURCES.txt @@ -1,70 +1,70 @@ .gitignore .pre-commit-config.yaml AUTHORS CODE_OF_CONDUCT.md CONTRIBUTORS LICENSE MANIFEST.in Makefile README.md conftest.py mypy.ini pyproject.toml pytest.ini requirements-swh.txt requirements-test.txt requirements.txt setup.cfg setup.py svn-lib-client-analysis.org tox.ini bin/init-svn-repository.sh bin/swh-svn docs/.gitignore docs/Makefile docs/conf.py docs/index.rst docs/swh-loader-svn.txt docs/_static/.placeholder docs/_templates/.placeholder install/install-pysvn.sh install/install-subvertpy.sh resources/svn.ini swh/__init__.py swh.loader.svn.egg-info/PKG-INFO swh.loader.svn.egg-info/SOURCES.txt swh.loader.svn.egg-info/dependency_links.txt swh.loader.svn.egg-info/entry_points.txt swh.loader.svn.egg-info/requires.txt swh.loader.svn.egg-info/top_level.txt swh/loader/__init__.py swh/loader/svn/__init__.py swh/loader/svn/converters.py swh/loader/svn/exception.py swh/loader/svn/loader.py swh/loader/svn/py.typed -swh/loader/svn/ra.py +swh/loader/svn/replay.py swh/loader/svn/svn.py swh/loader/svn/tasks.py swh/loader/svn/utils.py swh/loader/svn/tests/__init__.py swh/loader/svn/tests/conftest.py swh/loader/svn/tests/test_converters.py swh/loader/svn/tests/test_loader.org swh/loader/svn/tests/test_loader.py swh/loader/svn/tests/test_task.py swh/loader/svn/tests/test_utils.py swh/loader/svn/tests/data/httthttt.tgz swh/loader/svn/tests/data/mediawiki-repo-r407-eol-native-crlf.tgz swh/loader/svn/tests/data/penguinsdbtools2018.dump.gz swh/loader/svn/tests/data/pkg-doc-linux-r10.tgz swh/loader/svn/tests/data/pkg-doc-linux-r11.tgz swh/loader/svn/tests/data/pkg-doc-linux-r12.tgz swh/loader/svn/tests/data/pkg-gourmet-add-remove-dir.tgz swh/loader/svn/tests/data/pkg-gourmet-tampered-rev6-log.tgz swh/loader/svn/tests/data/pkg-gourmet-with-edge-case-links-and-files.tgz swh/loader/svn/tests/data/pkg-gourmet-with-external-id.tgz swh/loader/svn/tests/data/pkg-gourmet-with-updates.tgz swh/loader/svn/tests/data/pkg-gourmet-with-wrong-link-cases.tgz swh/loader/svn/tests/data/pkg-gourmet.tgz swh/loader/svn/tests/data/pyang-repo-r343-eol-native-mixed-lf-crlf.tgz \ No newline at end of file diff --git a/swh.loader.svn.egg-info/requires.txt b/swh.loader.svn.egg-info/requires.txt index 375430c..c46650b 100644 --- a/swh.loader.svn.egg-info/requires.txt +++ b/swh.loader.svn.egg-info/requires.txt @@ -1,17 +1,16 @@ click iso8601 -python-dateutil subvertpy>=0.9.4 typing-extensions swh.storage>=0.11.3 -swh.model>=0.4.0 +swh.model>=4.3.0 swh.scheduler>=0.0.39 swh.loader.core>=0.18 [testing] pytest pytest-mock pytest-postgresql swh.core[http]>=0.0.61 types-click types-python-dateutil diff --git a/swh/__init__.py b/swh/__init__.py index 8d9f151..b36383a 100644 --- a/swh/__init__.py +++ b/swh/__init__.py @@ -1,4 +1,3 @@ from pkgutil import extend_path -from typing import List -__path__: List[str] = extend_path(__path__, __name__) +__path__ = extend_path(__path__, __name__) diff --git a/swh/loader/__init__.py b/swh/loader/__init__.py index 8d9f151..b36383a 100644 --- a/swh/loader/__init__.py +++ b/swh/loader/__init__.py @@ -1,4 +1,3 @@ from pkgutil import extend_path -from typing import List -__path__: List[str] = extend_path(__path__, __name__) +__path__ = extend_path(__path__, __name__) diff --git a/swh/loader/svn/converters.py b/swh/loader/svn/converters.py index 3915140..cdef38b 100644 --- a/swh/loader/svn/converters.py +++ b/swh/loader/svn/converters.py @@ -1,82 +1,86 @@ # Copyright (C) 2015-2020 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 datetime from typing import Dict, Optional, Sequence, Tuple -from swh.model.model import Person, Revision, RevisionType, TimestampWithTimezone +import iso8601 -from .utils import strdate_to_timestamp +from swh.model.model import Person, Revision, RevisionType, TimestampWithTimezone -def svn_date_to_swh_date(strdate: Optional[str]) -> TimestampWithTimezone: +def svn_date_to_swh_date(strdate: Optional[bytes]) -> TimestampWithTimezone: """Convert a string date to an swh one. Args: - strdate: A string formatted for .utils.strdate_to_timestamp - to do its jobs + strdate: A string representing a date with format like + ``b'YYYY-mm-DDTHH:MM:SS.800722Z'`` Returns: An swh date format """ - return TimestampWithTimezone( - timestamp=strdate_to_timestamp(strdate), offset=0, negative_utc=False, - ) + if not strdate: # either None or empty string + dt = datetime.datetime(1970, 1, 1, tzinfo=datetime.timezone.utc) + else: + dt = iso8601.parse_date(strdate.decode("ascii")) + assert dt.tzinfo is not None, strdate + return TimestampWithTimezone.from_datetime(dt) def svn_author_to_swh_person(author: Optional[bytes]) -> Person: """Convert an svn author to an swh person. Default policy: No information is added. Args: author: the svn author (in bytes) Returns: a Person """ return Person.from_fullname(author or b"") def build_swh_revision( rev: int, commit: Dict, repo_uuid: bytes, dir_id: bytes, parents: Sequence[bytes] ) -> Revision: """Given a svn revision, build a swh revision. This adds an 'extra-headers' entry with the repository's uuid and the svn revision. Args: rev: the svn revision number commit: the commit data: revision id, date, author, and message repo_uuid: The repository's uuid dir_id: the tree's hash identifier parents: the revision's parents identifier Returns: The swh revision dictionary. """ author = commit["author_name"] msg = commit["message"] date = commit["author_date"] extra_headers: Tuple[Tuple[bytes, bytes], ...] = ( (b"svn_repo_uuid", repo_uuid), (b"svn_revision", str(rev).encode()), ) return Revision( type=RevisionType.SUBVERSION, date=date, committer_date=date, directory=dir_id, message=msg, author=author, committer=author, synthetic=True, extra_headers=extra_headers, parents=tuple(parents), ) diff --git a/swh/loader/svn/loader.py b/swh/loader/svn/loader.py index 0c414f8..4c93ed6 100644 --- a/swh/loader/svn/loader.py +++ b/swh/loader/svn/loader.py @@ -1,830 +1,833 @@ # Copyright (C) 2015-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information """Loader in charge of injecting either new or existing svn mirrors to swh-storage. """ from datetime import datetime from mmap import ACCESS_WRITE, mmap import os import pty import re import shutil from subprocess import Popen import tempfile from typing import Dict, Iterator, List, Optional, Sequence, Tuple from subvertpy import SubversionException from swh.loader.core.loader import BaseLoader from swh.loader.core.utils import clean_dangling_folders from swh.loader.exception import NotFound from swh.loader.svn.svn import SvnRepo from swh.model import from_disk, hashutil from swh.model.model import ( Content, Directory, Origin, Revision, SkippedContent, Snapshot, SnapshotBranch, TargetType, ) from swh.storage.algos.snapshot import snapshot_get_latest from swh.storage.interface import StorageInterface from . import converters from .exception import SvnLoaderHistoryAltered, SvnLoaderUneventful from .utils import ( OutputStream, init_svn_repo_from_archive_dump, init_svn_repo_from_dump, ) DEFAULT_BRANCH = b"HEAD" TEMPORARY_DIR_PREFIX_PATTERN = "swh.loader.svn." SUBVERSION_ERROR = re.compile(r".*(E[0-9]{6}):.*") SUBVERSION_NOT_FOUND = "E170013" class SvnLoader(BaseLoader): """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, max_content_size: Optional[int] = None, ): """Load a svn repository (either remote or local). Args: url: The default origin url origin_url: Optional original url override to use as origin reference in the archive. If not provided, "url" is used as origin. visit_date: Optional date to override the visit date incremental: If True, the default, starts from the last snapshot (if any). Otherwise, starts from the initial commit of the repository. temp_directory: The temporary directory to use as root directory for working directory computations debug: If true, run the loader in debug mode. At the end of the loading, the temporary working directory is not cleaned up to ease inspection. Defaults to false. check_revision: The number of svn commits between checks for hash divergence max_content_size: Default max content size allowed """ super().__init__( storage=storage, logging_class="swh.loader.svn.SvnLoader", max_content_size=max_content_size, ) # technical svn uri to act on svn repository self.svn_url = url # origin url as unique identifier for origin in swh archive self.origin_url = origin_url if origin_url else self.svn_url self.debug = debug self.temp_directory = temp_directory self.done = False self.svnrepo = None 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 self.incremental = incremental self.snapshot: Optional[Snapshot] = None # state from previous visit self.latest_snapshot = None self.latest_revision = None def pre_cleanup(self): """Cleanup potential dangling files from prior runs (e.g. OOM killed tasks) """ clean_dangling_folders( self.temp_directory, pattern_check=TEMPORARY_DIR_PREFIX_PATTERN, log=self.log, ) def cleanup(self): """Clean up the svn repository's working representation on disk. """ if not self.svnrepo: # could happen if `prepare` fails return if self.debug: self.log.error( """NOT FOR PRODUCTION - debug flag activated Local repository not cleaned up for investigation: %s""", self.svnrepo.local_url.decode("utf-8"), ) return self.svnrepo.clean_fs() def swh_revision_hash_tree_at_svn_revision(self, revision: 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) h = from_disk.Directory.from_disk(path=local_url).hash self.svnrepo.clean_fs(local_dirname) return h def _latest_snapshot_revision( self, origin_url: str, ) -> Optional[Tuple[Snapshot, Revision]]: """Look for latest snapshot revision and returns it if any. Args: origin_url: Origin identifier previous_swh_revision: possible previous swh revision (either a dict or revision identifier) Returns: Tuple of the latest Snapshot from the previous visit and its targeted revision if any or None otherwise. """ storage = self.storage latest_snapshot = snapshot_get_latest(storage, origin_url) if not latest_snapshot: return None branches = latest_snapshot.branches if not branches: return None branch = branches.get(DEFAULT_BRANCH) if not branch: return None if branch.target_type != TargetType.REVISION: return None swh_id = branch.target revision = storage.revision_get([swh_id])[0] if not revision: return None return latest_snapshot, revision def build_swh_revision( self, rev: 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 - hash_data_per_revs = self.svnrepo.swh_hash_data_at_revision(revision_start) - - rev = revision_start - commit, root_dir = list(hash_data_per_revs)[0] + commit, root_dir = self.svnrepo.swh_hash_data_at_revision(revision_start) dir_id = root_dir.hash - swh_revision = self.build_swh_revision(rev, commit, dir_id, parents) + 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, Dict[int, Tuple[bytes, ...]]]: """Determine from where to start the loading. Returns: tuple (revision_start, revision_end, revision_parents) Raises: SvnLoaderHistoryAltered: When a hash divergence has been detected (should not happen) SvnLoaderUneventful: Nothing changed since last visit """ assert self.svnrepo is not None, "svnrepo initialized in the `prepare` method" revision_head = self.svnrepo.head_revision() if revision_head == 0: # empty repository case revision_start = 0 revision_end = 0 else: # default configuration revision_start = self.svnrepo.initial_revision() revision_end = revision_head revision_parents: Dict[int, Tuple[bytes, ...]] = {revision_start: ()} # start from a previous revision if any if self.incremental and self.latest_revision is not None: extra_headers = dict(self.latest_revision.extra_headers) revision_start = int(extra_headers[b"svn_revision"]) revision_parents = { revision_start: self.latest_revision.parents, } self.log.debug( "svn export --ignore-keywords %s@%s", self.svnrepo.remote_url, revision_start, ) if not self.check_history_not_altered(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 # and the parent become the latest know revision for # that repository revision_parents[revision_start] = (self.latest_revision.id,) if revision_start > revision_end: msg = "%s@%s already injected." % (self.svnrepo.remote_url, revision_end) raise SvnLoaderUneventful(msg) self.log.info( "Processing revisions [%s-%s] for %s", revision_start, revision_end, self.svnrepo, ) return revision_start, revision_end, revision_parents def _check_revision_divergence(self, 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, revision_parents ) -> Iterator[ Tuple[List[Content], List[SkippedContent], List[Directory], Revision] ]: """Process svn revisions from revision_start to revision_end. At each svn revision, apply new diffs and simultaneously compute swh hashes. This yields those computed swh hashes as a tuple (contents, directories, revision). Note that at every `self.check_revision`, a supplementary check takes place to check for hash-tree divergence (related T570). Yields: tuple (contents, directories, revision) of dict as a dictionary with keys, sha1_git, sha1, etc... Raises: ValueError in case of a hash divergence detection """ gen_revs = svnrepo.swh_hash_data_per_revision(revision_start, revision_end) swh_revision = None count = 0 for rev, nextrev, commit, new_objects, root_directory in gen_revs: count += 1 # Send the associated contents/directories _contents, _skipped_contents, _directories = new_objects # compute the fs tree's checksums dir_id = root_directory.hash swh_revision = self.build_swh_revision( rev, commit, dir_id, revision_parents.get(rev, ()) ) self.log.debug( "rev: %s, swhrev: %s, dir: %s", rev, hashutil.hash_to_hex(swh_revision.id), hashutil.hash_to_hex(dir_id), ) if ( self.check_revision and self.check_revision != 0 and count % self.check_revision == 0 ): self._check_revision_divergence(rev, dir_id) if nextrev: revision_parents[nextrev] = [swh_revision.id] yield _contents, _skipped_contents, _directories, swh_revision + 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 prepare_origin_visit(self): self.origin = Origin(url=self.origin_url if self.origin_url else self.svn_url) def prepare(self): latest_snapshot_revision = self._latest_snapshot_revision(self.origin_url) if latest_snapshot_revision: self.latest_snapshot, self.latest_revision = latest_snapshot_revision + self._snapshot = self.latest_snapshot + self._last_revision = self.latest_revision local_dirname = self._create_tmp_dir(self.temp_directory) try: self.svnrepo = SvnRepo( self.svn_url, self.origin_url, local_dirname, self.max_content_size ) except SubversionException as e: error_msgs = [ "Unable to connect to a repository at URL", "Unknown URL type", ] for msg in error_msgs: if msg in e.args[0]: self._load_status = "uneventful" raise NotFound(e) raise try: revision_start, revision_end, revision_parents = self.start_from() self.swh_revision_gen = self.process_svn_revisions( self.svnrepo, revision_start, revision_end, revision_parents ) except SvnLoaderUneventful as e: self.log.warning(e) - if self.latest_snapshot: - self._snapshot = self.latest_snapshot self.done = True self._load_status = "uneventful" except SvnLoaderHistoryAltered as e: self.log.error(e) self.done = True self._visit_status = "partial" def fetch_data(self): """Fetching svn revision information. This will apply svn revision as patch on disk, and at the same time, compute the swh hashes. In effect, fetch_data fetches those data and compute the necessary swh objects. It's then stored in the internal state instance variables (initialized in `_prepare_state`). This is up to `store_data` to actually discuss with the storage to store those objects. Returns: bool: True to continue fetching data (next svn revision), False to stop. """ - data = None + if self.done: return False try: data = next(self.swh_revision_gen) self._load_status = "eventful" except StopIteration: - self.done = True + self.done = True # Stopping iteration self._visit_status = "full" - return False # Stopping iteration except Exception as e: # svn:external, hash divergence, i/o error... self.log.exception(e) - self.done = True + self.done = True # Stopping iteration self._visit_status = "partial" - return False # Stopping iteration - self._contents, self._skipped_contents, self._directories, rev = data - if rev: - self._last_revision = rev - self._revisions.append(rev) - return True # next svn revision + 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, max_content_size: Optional[int] = None, ): 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, max_content_size=max_content_size, ) self.archive_path = archive_path self.temp_dir = None self.repo_path = None def prepare(self): self.log.info("Archive to mount and load %s", self.archive_path) self.temp_dir, self.repo_path = init_svn_repo_from_archive_dump( self.archive_path, prefix=TEMPORARY_DIR_PREFIX_PATTERN, suffix="-%s" % os.getpid(), root_dir=self.temp_directory, ) self.svn_url = f"file://{self.repo_path}" super().prepare() def cleanup(self): super().cleanup() if self.temp_dir and os.path.exists(self.temp_dir): self.log.debug( "Clean up temporary directory dump %s for project %s", self.temp_dir, os.path.basename(self.repo_path), ) shutil.rmtree(self.temp_dir) class SvnLoaderFromRemoteDump(SvnLoader): """Create a subversion repository dump 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, max_content_size: Optional[int] = None, ): 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, max_content_size=max_content_size, ) 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 - if len(stderr_lines) > 1: - last_dumped_rev = stderr_lines[-2] + 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: raise Exception( ( "Last dumped subversion revision (%s) is " "lesser than the last one loaded into the " "archive (%s)." ) % (last_dumped_rev, last_loaded_svn_rev) ) if SUBVERSION_NOT_FOUND in error_codes: raise NotFound( f"{SUBVERSION_NOT_FOUND}: Repository never existed or disappeared" ) raise Exception( "An error occurred when running svnrdump and " "no exploitable dump file has been generated.\n" + "\n".join(error_messages) ) def prepare(self): # First, check if previous revisions have been loaded for the # subversion origin and get the number of the last one last_loaded_svn_rev = self.get_last_loaded_svn_rev(self.svn_url) # Then 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 = SvnRepo( 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/ra.py b/swh/loader/svn/ra.py deleted file mode 100644 index 91ead84..0000000 --- a/swh/loader/svn/ra.py +++ /dev/null @@ -1,599 +0,0 @@ -# Copyright (C) 2016-2021 The Software Heritage developers -# See the AUTHORS file at the top-level directory of this distribution -# License: GNU General Public License version 3, or any later version -# See top-level LICENSE file for more information - -"""Remote Access client to svn server. - -""" - -import codecs -import dataclasses -import os -import shutil -import tempfile -from typing import Dict, List, Optional, Tuple - -import click -from subvertpy import delta, properties -from subvertpy.ra import Auth, RemoteAccess, get_username_provider - -from swh.model import from_disk, hashutil -from swh.model.model import Content, Directory, SkippedContent - -_eol_style = {"native": b"\n", "CRLF": b"\r\n", "LF": b"\n", "CR": b"\r"} - - -def _normalize_line_endings(lines, eol_style="native"): - r"""Normalize line endings to unix (\\n), windows (\\r\\n) or mac (\\r). - - Args: - lines (bytes): The lines to normalize - - line_ending (str): The line ending format as defined for - svn:eol-style property. Acceptable values are 'native', - 'CRLF', 'LF' and 'CR' - - Returns: - bytes: 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, target_stream): - """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, sbuf=sbuf, target_stream=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): - """Read the svn link's content. - - Args: - data (bytes): 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) - src = split_byte.join(src) - return filetype, src - - -def is_file_an_svnlink_p(fullpath): - """Determine if a filepath is an svnlink or something else. - - Args: - fullpath (str/bytes): Full path to the potential symlink to check - - Returns: - boolean value to determine if it's indeed a symlink (as per - svn) or not. - - """ - with open(fullpath, "rb") as f: - filetype, src = read_svn_link(f.read()) - return filetype == b"link", src - - -def _ra_codecs_error_handler(e): - """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 (UnicodeDecodeError): exception raised during the svn - properties decoding. - - """ - return "", e.end - - -DEFAULT_FLAG = 0 -EXEC_FLAG = 1 -NOEXEC_FLAG = 2 - -SVN_PROPERTY_EOL = "svn:eol-style" - - -@dataclasses.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", - ] - - def __init__(self, directory, rootpath, path, state: FileState): - self.directory = directory - self.path = path - self.fullpath = os.path.join(rootpath, path) - self.state = state - - def change_prop(self, key, value): - 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): - """Convert the svnlink to a symlink on disk. - - This function expects self.fullpath to be a svn link. - - Args: - src (bytes): 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): - """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): - 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): - """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.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) - else: # not a real link ... - # when a file with the svn:special property set is not a svn link, - # the svn export operation will extract a truncated version of that file - # if it contains a null byte (see create_special_file_from_stream - # implementation in libsvn_subr/subst.c), so ensure to produce the - # same file as the export operation. - with open(self.fullpath, "rb") as f: - content = f.read() - with open(self.fullpath, "wb") as f: - exported_data = content.split(b"\x00")[0] - 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 - f.write(exported_data) - 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) - - -class BaseDirEditor: - """Base class implementation of dir editor. - - see :class:`DirEditor` for an implementation that hashes every - directory encountered. - - Instantiate a new class inheriting from this class and define the following - functions:: - - def update_checksum(self): - # Compute the checksums at current state - - def open_directory(self, *args): - # Update an existing folder. - - def add_directory(self, *args): - # Add a new one. - - """ - - __slots__ = ["directory", "rootpath"] - - def __init__(self, directory, rootpath, file_states: Dict[str, FileState]): - self.directory = directory - self.rootpath = rootpath - # build directory on init - os.makedirs(rootpath, exist_ok=True) - self.file_states = file_states - - def remove_child(self, path): - """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. - - """ - try: - entry_removed = self.directory[path] - except KeyError: - entry_removed = None - else: - 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 update_checksum(self): - raise NotImplementedError("This should be implemented.") - - def open_directory(self, *args): - raise NotImplementedError("This should be implemented.") - - def add_directory(self, *args): - raise NotImplementedError("This should be implemented.") - - def open_file(self, *args): - """Updating existing file. - - """ - path = os.fsencode(args[0]) - self.directory[path] = from_disk.Content() - fullpath = os.path.join(self.rootpath, path) - return FileEditor( - self.directory, - rootpath=self.rootpath, - path=path, - state=self.file_states[fullpath], - ) - - def add_file(self, path, copyfrom_path=None, copyfrom_rev=-1): - """Creating a new file. - - """ - path = os.fsencode(path) - self.directory[path] = from_disk.Content() - fullpath = os.path.join(self.rootpath, path) - self.file_states[fullpath] = FileState() - return FileEditor( - self.directory, self.rootpath, path, state=self.file_states[fullpath] - ) - - def change_prop(self, key, value): - """Change property callback on directory. - - """ - if key == properties.PROP_EXTERNALS: - raise ValueError("Property '%s' detected. Not implemented yet." % key) - - def delete_entry(self, path, revision): - """Remove a path. - - """ - fullpath = os.path.join(self.rootpath, path.encode("utf-8")) - self.file_states.pop(fullpath, None) - self.remove_child(path.encode("utf-8")) - - def close(self): - """Function called when we finish walking a repository. - - """ - self.update_checksum() - - -class DirEditor(BaseDirEditor): - """Directory Editor in charge of updating directory hashes computation. - - This implementation includes empty folder in the hash computation. - - """ - - def update_checksum(self): - """Update the root path self.path's checksums according to the - children's objects. - - This function is expected to be called when the folder has - been completely 'walked'. - - """ - pass - - def open_directory(self, *args): - """Updating existing directory. - - """ - return self - - def add_directory(self, path, copyfrom_path=None, copyfrom_rev=-1): - """Adding a new directory. - - """ - path = os.fsencode(path) - os.makedirs(os.path.join(self.rootpath, path), exist_ok=True) - self.directory[path] = from_disk.Directory() - return self - - -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, directory): - self.rootpath = rootpath - self.directory = directory - self.file_states: Dict[str, FileState] = {} - - def set_target_revision(self, revnum): - pass - - def abort(self): - pass - - def close(self): - pass - - def open_root(self, base_revnum): - return DirEditor( - self.directory, rootpath=self.rootpath, file_states=self.file_states - ) - - -class Replay: - """Replay class. - """ - - def __init__(self, conn, rootpath, directory=None): - self.conn = conn - self.rootpath = rootpath - if directory is None: - directory = from_disk.Directory() - self.directory = directory - self.editor = Editor(rootpath=rootpath, directory=directory) - - def replay(self, rev): - """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, rev + 1, self.editor) - codecs.register_error("strict", codecs.strict_errors) - return self.editor.directory - - def compute_objects( - self, rev: int - ) -> Tuple[List[Content], List[SkippedContent], List[Directory]]: - """Compute objects 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) - return from_disk.iter_directory(self.directory) - - -@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/replay.py b/swh/loader/svn/replay.py new file mode 100644 index 0000000..6078f50 --- /dev/null +++ b/swh/loader/svn/replay.py @@ -0,0 +1,906 @@ +# 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 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 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 + + self.editor.modified_paths.add(path) + + def change_prop(self, key: str, value: str) -> None: + 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 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.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.client.export( + os.path.join(self.svnrepo.remote_url.encode(), self.path), + to=self.fullpath, + 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) + + +@dataclass +class DirState: + """Persists some directory states (eg. externals) across revisions while + replaying them.""" + + externals: Dict[str, Tuple[str, Optional[int]]] = field(default_factory=dict) + + +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, Tuple[str, Optional[int], bool]] = {} + + # repository root dir has empty path + if path: + self.editor.modified_paths.add(path) + + 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. + + """ + try: + entry_removed = self.directory[path] + except KeyError: + entry_removed = None + else: + 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] + + self.editor.modified_paths.discard(path) + + def open_directory(self, path: str, *args) -> DirEditor: + """Updating existing directory. + + """ + 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, *args) -> DirEditor: + """Adding a new directory. + + """ + path_bytes = os.fsencode(path) + + os.makedirs(os.path.join(self.rootpath, path_bytes), exist_ok=True) + if path_bytes and path_bytes not in self.directory: + self.dir_states[path_bytes] = DirState() + self.directory[path_bytes] = from_disk.Directory() + + return DirEditor( + self.directory, + self.rootpath, + path_bytes, + self.file_states, + self.dir_states, + svnrepo=self.svnrepo, + ) + + def open_file(self, *args) -> FileEditor: + """Updating existing file. + + """ + path = os.fsencode(args[0]) + self.directory[path] = from_disk.Content() + fullpath = os.path.join(self.rootpath, path) + return FileEditor( + self.directory, + rootpath=self.rootpath, + path=path, + state=self.file_states[fullpath], + svnrepo=self.svnrepo, + ) + + def add_file(self, path: str, *args) -> FileEditor: + """Creating a new file. + + """ + path_bytes = os.fsencode(path) + self.directory[path_bytes] = from_disk.Content() + fullpath = os.path.join(self.rootpath, path_bytes) + self.file_states[fullpath] = FileState() + 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 = {} + if value is not None: + # 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] = (external_url, revision, relative_url) + + 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. + + """ + path_bytes = os.fsencode(path) + if path_bytes not in self.editor.external_paths: + fullpath = os.path.join(self.rootpath, path_bytes) + self.file_states.pop(fullpath, None) + self.remove_child(path_bytes) + + def close(self): + """Function called when we finish processing a repository. + + SVN external definitions are processed by it. + """ + + 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 + old_externals = set(prev_externals) - set(self.externals) + for old_external in old_externals: + self.remove_external_path(os.fsencode(old_external)) + 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, (external_url, revision, relative_url) in externals.items(): + external = (external_url, revision) + dest_path = os.fsencode(path) + dest_fullpath = os.path.join(self.path, dest_path) + if ( + path in prev_externals + and prev_externals[path] == external + and dest_fullpath in self.directory + ): + # external already exported, nothing to do + continue + + 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: + logger.debug( + "Exporting external %s to path %s", external_url, path + ) + self.svnrepo.client.export( + external_url.rstrip("/"), + to=temp_path, + rev=revision, + ignore_keywords=True, + ) + self.editor.externals_cache[external] = temp_path + self.editor.valid_externals[dest_fullpath] = ( + external_url, + relative_url, + ) + + 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(current_path) + for subpath in dest_path_split[:-1]: + current_path = os.path.join(current_path, subpath) + self.add_directory(current_path) + + if os.path.exists(temp_path): + # external successfully exported + + # remove previous path in from_disk model + self.remove_child(dest_fullpath) + # copy exported path to reconstructed filesystem + fullpath = os.path.join(self.rootpath, dest_fullpath) + + self.editor.external_paths.add(dest_fullpath) + self.editor.modified_paths.add(dest_fullpath) + # update from_disk model and store external paths + self.editor.external_paths.add(dest_fullpath) + if os.path.isfile(temp_path): + shutil.copy(temp_path, fullpath) + self.directory[dest_fullpath] = from_disk.Content.from_file( + path=fullpath + ) + else: + shutil.copytree(temp_path, fullpath, symlinks=True) + self.directory[dest_fullpath] = from_disk.Directory.from_disk( + path=fullpath + ) + external_paths = set() + for root, dirs, files in os.walk(fullpath): + external_paths.update( + [ + os.path.join(root.replace(self.rootpath + b"/", b""), p) + for p in chain(dirs, files) + ] + ) + self.editor.external_paths.update(external_paths) + self.editor.modified_paths.update(external_paths) + + # ensure hash update for the directory with externals set + self.directory[self.path].update_hash(force=True) + + # backup externals in directory state + if self.externals: + self.dir_states[self.path].externals = self.externals + + self.svnrepo.has_relative_externals = any( + [relative_url for (_, relative_url) in self.editor.valid_externals.values()] + ) + + def remove_external_path(self, external_path: bytes) -> None: + """Remove a previously exported SVN external path from + the reconstruted filesystem. + """ + fullpath = os.path.join(self.path, external_path) + self.remove_child(fullpath) + self.editor.external_paths.discard(fullpath) + self.editor.valid_externals.pop(fullpath, None) + for path in list(self.editor.external_paths): + if path.startswith(fullpath + b"/"): + self.editor.external_paths.remove(path) + subpath_split = external_path.split(b"/")[:-1] + for i in reversed(range(1, len(subpath_split) + 1)): + # delete external sub-directory only if it is empty + subpath = os.path.join(self.path, b"/".join(subpath_split[0:i])) + if not os.listdir(os.path.join(self.rootpath, subpath)): + 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, + rev=self.editor.revnum, + ignore_keywords=True, + ) + if os.path.isfile(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, + ): + 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: Set[bytes] = set() + 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[Tuple[str, Optional[int]], str] = {} + self.svnrepo = svnrepo + self.revnum = None + # to store the set of paths added or modified when replaying a revision + self.modified_paths: Set[bytes] = set() + + 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: + # a new revision is being replayed so clear the modified_paths set + self.modified_paths.clear() + 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, + ): + 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 + ) + + def replay(self, rev: 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, rev + 1, self.editor) + codecs.register_error("strict", codecs.strict_errors) + return self.editor.directory + + def compute_objects( + self, rev: 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) + + contents: List[Content] = [] + skipped_contents: List[SkippedContent] = [] + directories: List[Directory] = [] + + directories.append(self.editor.directory.to_model()) + for path in self.editor.modified_paths: + obj = self.directory[path].to_model() + 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) + + 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 3e1ee5d..fdfc5da 100644 --- a/swh/loader/svn/svn.py +++ b/swh/loader/svn/svn.py @@ -1,296 +1,344 @@ -# Copyright (C) 2015-2021 The Software Heritage developers +# 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 client, properties 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, ra +from . import converters, replay +from .utils import 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, ): self.remote_url = remote_url.rstrip("/") self.origin_url = origin_url auth = Auth([get_username_provider()]) # one connection for log iteration self.conn_log = RemoteAccess(self.remote_url, auth=auth) # another for replay self.conn = RemoteAccess(self.remote_url, auth=auth) # one client for update operation self.client = client.Client(auth=auth) self.local_dirname = local_dirname local_name = os.path.basename(self.remote_url) self.local_url = os.path.join(self.local_dirname, local_name).encode("utf-8") self.uuid = self.conn.get_uuid().encode("utf-8") - self.swhreplay = ra.Replay(conn=self.conn, rootpath=self.local_url) + self.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.replay_started = False 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: str) -> TimestampWithTimezone: + 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) ) return { "rev": rev, "author_date": author_date, "author_name": author, "message": message, } 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=False, ): yield self.__to_entry(log_entry) 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 oustide 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 and self.remote_url.startswith("file://"): + # 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 + with tempfile.TemporaryDirectory( + dir=self.local_dirname, prefix=f"checkout-revision-{revision}." + ) as co_dirname: + self.client.checkout( + self.remote_url, co_dirname, revision, ignore_externals=True + ) + # get all svn:externals properties recursively + externals = self.client.propget( + "svn:externals", co_dirname, None, revision, True + ) + self.has_relative_externals = False + for path, external_defs in externals.items(): + if self.has_relative_externals: + break + for external_def in os.fsdecode(external_defs).split("\n"): + # skip empty line or comment + if not external_def or external_def.startswith("#"): + continue + _, _, _, relative_url = parse_external_definition( + external_def.rstrip("\r"), path, self.origin_url + ) + if relative_url: + self.has_relative_externals = True + url = self.origin_url + break + self.client.export( - self.remote_url, to=local_url, rev=revision, ignore_keywords=True + url.rstrip("/"), to=local_url, rev=revision, ignore_keywords=True, ) return local_dirname, os.fsencode(local_url) def swh_hash_data_per_revision( self, start_revision: int, end_revision: int ) -> Iterator[ Tuple[ int, Optional[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 - nextrev: next revision or None if we reached end_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 == end_revision: nextrev = None else: nextrev = rev + 1 if rev >= start_revision: # start yielding new data to archive once we reached the revision to # resume the loading from yield rev, nextrev, commit, objects, self.swhreplay.directory def swh_hash_data_at_revision( self, revision: int - ) -> Iterator[Tuple[Dict, DirectoryFromDisk]]: + ) -> 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) - yield commit, directory + 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_converters.py b/swh/loader/svn/tests/test_converters.py index 7dd7696..2021918 100644 --- a/swh/loader/svn/tests/test_converters.py +++ b/swh/loader/svn/tests/test_converters.py @@ -1,124 +1,120 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from swh.loader.svn import converters from swh.model.hashutil import hash_to_bytes from swh.model.model import Person, Revision, Timestamp, TimestampWithTimezone def test_svn_author_to_swh_person(): """The author should have name, email and fullname filled. """ actual_person = converters.svn_author_to_swh_person(b"tony ") assert actual_person == Person.from_dict( {"fullname": b"tony ", "name": b"tony", "email": b"ynot@dagobah",} ) def test_svn_author_to_swh_person_no_email(): """The author and fullname should be the same as the input (author). """ actual_person = converters.svn_author_to_swh_person(b"tony") assert actual_person == Person.from_dict( {"fullname": b"tony", "name": b"tony", "email": None,} ) def test_svn_author_to_swh_person_empty_person(): """Empty person has only its fullname filled with the empty byte-string. """ actual_person = converters.svn_author_to_swh_person(b"") assert actual_person == Person.from_dict( {"fullname": b"", "name": None, "email": None,} ) def test_build_swh_revision_default(): """This should build the swh revision with the swh revision's extra headers about the repository. """ dir_id = hash_to_bytes("d6e08e19159f77983242877c373c75222d5ae9dd") date = TimestampWithTimezone( - timestamp=Timestamp(seconds=1088108379, microseconds=0), - offset=0, - negative_utc=False, + timestamp=Timestamp(seconds=1088108379, microseconds=0), offset_bytes=b"+0000" ) actual_rev = converters.build_swh_revision( repo_uuid=b"uuid", dir_id=dir_id, commit={ "author_name": Person( name=b"theo", email=b"theo@uuid", fullname=b"theo " ), "message": b"commit message", "author_date": date, }, rev=10, parents=(), ) expected_rev = Revision.from_dict( { "date": date.to_dict(), "committer_date": date.to_dict(), "type": "svn", "directory": dir_id, "message": b"commit message", "author": { "name": b"theo", "email": b"theo@uuid", "fullname": b"theo ", }, "committer": { "name": b"theo", "email": b"theo@uuid", "fullname": b"theo ", }, "synthetic": True, "extra_headers": ((b"svn_repo_uuid", b"uuid"), (b"svn_revision", b"10"),), "parents": (), } ) assert actual_rev == expected_rev def test_svn_date_to_swh_date(): """The timestamp should not be tampered with and include the decimals. """ assert converters.svn_date_to_swh_date( - "2011-05-31T06:04:39.500900Z" + b"2011-05-31T06:04:39.500900Z" ) == TimestampWithTimezone( timestamp=Timestamp(seconds=1306821879, microseconds=500900), - offset=0, - negative_utc=False, + offset_bytes=b"+0000", ) assert converters.svn_date_to_swh_date( - "2011-05-31T06:04:39.800722Z" + b"2011-05-31T06:04:39.800722Z" ) == TimestampWithTimezone( timestamp=Timestamp(seconds=1306821879, microseconds=800722), - offset=0, - negative_utc=False, + offset_bytes=b"+0000", ) def test_svn_date_to_swh_date_epoch(): """Empty date should be EPOCH (timestamp and offset at 0).""" # It should return 0, epoch default_tstz = TimestampWithTimezone( - timestamp=Timestamp(seconds=0, microseconds=0), offset=0, negative_utc=False, + timestamp=Timestamp(seconds=0, microseconds=0), offset_bytes=b"+0000" ) assert converters.svn_date_to_swh_date("") == default_tstz assert converters.svn_date_to_swh_date(None) == default_tstz diff --git a/swh/loader/svn/tests/test_loader.py b/swh/loader/svn/tests/test_loader.py index 3ad5a44..c2f73f7 100644 --- a/swh/loader/svn/tests/test_loader.py +++ b/swh/loader/svn/tests/test_loader.py @@ -1,1780 +1,2525 @@ -# Copyright (C) 2016-2021 The Software Heritage developers +# 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 + from enum import Enum from io import BytesIO import os import shutil import subprocess from typing import Any, Dict, List import pytest from subvertpy import SubversionException, delta, repos from subvertpy.ra import Auth, RemoteAccess, get_username_provider from typing_extensions import TypedDict from swh.loader.svn.loader import ( SvnLoader, SvnLoaderFromDumpArchive, SvnLoaderFromRemoteDump, ) from swh.loader.svn.svn import SvnRepo -from swh.loader.svn.utils import init_svn_repo_from_dump +from swh.loader.svn.utils import init_svn_repo_from_dump, svn_urljoin from swh.loader.tests import ( assert_last_visit_matches, check_snapshot, get_stats, prepare_repository_from_archive, ) from swh.model.from_disk import DentryPerms from swh.model.hashutil import hash_to_bytes from swh.model.model import Snapshot, SnapshotBranch, TargetType GOURMET_SNAPSHOT = Snapshot( id=hash_to_bytes("889cacc2731e3312abfb2b1a0c18ade82a949e07"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("4876cb10aec6f708f7466dddf547567b65f6c39c"), target_type=TargetType.REVISION, ) }, ) GOURMET_UPDATES_SNAPSHOT = Snapshot( id=hash_to_bytes("11086d15317014e43d2438b7ffc712c44f1b8afe"), branches={ b"HEAD": SnapshotBranch( target=hash_to_bytes("171dc35522bfd17dda4e90a542a0377fb2fc707a"), target_type=TargetType.REVISION, ) }, ) def test_loader_svn_not_found_no_mock(swh_storage, tmp_path): """Given an unknown repository, the loader visit ends up in status not_found""" repo_url = "unknown-repository" loader = SvnLoader(swh_storage, repo_url, 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_external_properties(swh_storage, datadir, tmp_path): - """Repository with svn:external properties cannot be fully ingested yet - - """ - archive_name = "pkg-gourmet" - archive_path = os.path.join(datadir, "pkg-gourmet-with-external-id.tgz") - repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) - - loader = SvnLoader(swh_storage, repo_url, temp_directory=tmp_path) - - assert loader.load() == {"status": "eventful"} - gourmet_externals_snapshot = Snapshot( - id=hash_to_bytes("19cb68d0a3f22372e2b7017ea5e2a2ea5ae3e09a"), - branches={ - b"HEAD": SnapshotBranch( - target=hash_to_bytes("82a7a4a09f9549223429143ba36ad77375e33c5c"), - target_type=TargetType.REVISION, - ) - }, - ) - check_snapshot(gourmet_externals_snapshot, loader.storage) - assert_last_visit_matches( - loader.storage, - repo_url, - status="partial", - type="svn", - snapshot=gourmet_externals_snapshot.id, - ) - - stats = get_stats(loader.storage) - assert stats["origin"] == 1 - assert stats["origin_visit"] == 1 - assert stats["snapshot"] == 1 - # repository holds 21 revisions, but the last commit holds an 'svn:externals' - # property which will make the loader-svn stops at the last revision prior to the - # bad one - assert stats["revision"] == 21 - 1 # commit with the svn:external property - - def test_loader_svn_with_symlink(swh_storage, datadir, tmp_path): """Repository with symlinks should be ingested ok Edge case: - first create a file and commit it. Remove it, then add folder holding the same name, commit. - do the same scenario with symbolic link (instead of file) """ archive_name = "pkg-gourmet" archive_path = os.path.join( datadir, "pkg-gourmet-with-edge-case-links-and-files.tgz" ) repo_url = prepare_repository_from_archive(archive_path, archive_name, tmp_path) loader = SvnLoader(swh_storage, repo_url, 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, 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) 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, ) origin_url = repo_url + "2" # rename to another origin 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) origin_url = f"svn://{archive_name}" dump_filename = f"{archive_name}.dump" with open(os.path.join(tmp_path, dump_filename), "wb") as dump_file: # create compressed dump file of pkg-gourmet repo subprocess.run(["svnrdump", "dump", repo_url], stdout=dump_file) subprocess.run(["gzip", dump_filename], cwd=tmp_path) # load svn repo from that compressed dump file loader = SvnLoaderFromDumpArchive( swh_storage, url=origin_url, archive_path=os.path.join(tmp_path, f"{dump_filename}.gz"), temp_directory=tmp_path, ) assert loader.load() == {"status": "eventful"} assert_last_visit_matches( loader.storage, origin_url, status="full", type="svn", snapshot=GOURMET_SNAPSHOT.id, ) check_snapshot(GOURMET_SNAPSHOT, loader.storage) assert get_stats(loader.storage) == { "content": 19, "directory": 17, "origin": 1, "origin_visit": 1, "release": 0, "revision": 6, "skipped_content": 0, "snapshot": 1, } class CommitChangeType(Enum): AddOrUpdate = 1 Delete = 2 class CommitChange(TypedDict, total=False): change_type: CommitChangeType path: str properties: Dict[str, str] data: bytes def add_commit(repo_url: str, message: str, changes: List[CommitChange]) -> None: conn = RemoteAccess(repo_url, auth=Auth([get_username_provider()])) editor = conn.get_commit_editor({"svn:log": message}) root = editor.open_root() for change in changes: if change["change_type"] == CommitChangeType.Delete: root.delete_entry(change["path"].rstrip("/")) else: dir_change = change["path"].endswith("/") split_path = change["path"].rstrip("/").split("/") for i in range(len(split_path)): path = "/".join(split_path[0 : i + 1]) if i < len(split_path) - 1: try: root.add_directory(path).close() except SubversionException: pass else: if dir_change: - root.add_directory(path).close() + try: + dir = root.add_directory(path) + except SubversionException: + dir = root.open_directory(path) + if "properties" in change: + for prop, value in change["properties"].items(): + dir.change_prop(prop, value) + dir.close() else: try: file = root.add_file(path) except SubversionException: file = root.open_file(path) if "properties" in change: for prop, value in change["properties"].items(): file.change_prop(prop, value) if "data" in change: txdelta = file.apply_textdelta() delta.send_stream(BytesIO(change["data"]), txdelta) file.close() root.close() editor.close() -def test_loader_eol_style_file_property_handling_edge_case(swh_storage, tmp_path): - # create a repository +def create_repo(tmp_path): repo_path = os.path.join(tmp_path, "tmprepo") repos.create(repo_path) repo_url = f"file://{repo_path}" + return repo_url + + +@pytest.fixture +def repo_url(tmpdir_factory): + # create a repository + return create_repo(tmpdir_factory.mktemp("repos")) + + +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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path): +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) - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" for filename in ("foo", "bar", "baz"): add_commit( repo_url, f"Add {filename} file", [ CommitChange( change_type=CommitChangeType.AddOrUpdate, path=filename, data=f"{filename}\n".encode(), ) ], ) loader = SvnLoader(swh_storage, repo_url, 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_with_null_byte( - swh_storage, tmp_path -): +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 will be truncated when performing an export operation if it contains - a null byte. Indeed, subversion will treat the file content as text but - it might be a binary file containing null bytes.""" - - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" + 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", + ( + "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, - "Remove the svn:special property on the previously added file", + ( + "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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path + swh_storage, mocker, repo_url, tmp_path ): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" # first commit add_commit( repo_url, ("Add a 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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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, tmp_path): - # create a repository - repo_path = os.path.join(tmp_path, "tmprepo") - repos.create(repo_path) - repo_url = f"file://{repo_path}" +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 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.fixture +def external_repo_url(tmpdir_factory): + # create a repository + return create_repo(tmpdir_factory.mktemp("external")) + + +def test_loader_with_valid_svn_externals( + swh_storage, repo_url, external_repo_url, tmp_path +): + # first commit on external + add_commit( + external_repo_url, + "Create some directories and files in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/hello/hello-world", + properties={"svn:executable": "*"}, + data=b"#!/bin/bash\necho Hello World !", + ), + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="foo.sh", + properties={"svn:executable": "*"}, + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + # first commit + add_commit( + repo_url, + "Create repository structure.", + [ + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="branches/",), + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="tags/",), + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="trunk/",), + ], + ) + + # second commit + add_commit( + repo_url, + ( + "Set svn:externals property on trunk/externals path of repository to load." + "One external targets a remote directory and another one a remote file." + ), + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/externals/", + properties={ + "svn:externals": ( + f"{svn_urljoin(external_repo_url, 'code/hello')} hello\n" + f"{svn_urljoin(external_repo_url, 'foo.sh')} foo.sh" + ) + }, + ), + ], + ) + + # 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) + + # third commit + add_commit( + repo_url, + "Unset svn:externals property on trunk/externals path", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/externals/", + properties={"svn:externals": None}, + ), + ], + ) + + # second 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) + + +def test_loader_with_invalid_svn_externals(swh_storage, repo_url, tmp_path): + + # first commit + add_commit( + repo_url, + "Create repository structure.", + [ + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="branches/",), + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="tags/",), + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="trunk/",), + ], + ) + + # second commit + add_commit( + repo_url, + ( + "Set svn:externals property on trunk/externals path of repository to load." + "The externals URLs are not valid." + ), + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/externals/", + properties={ + "svn:externals": ( + "file:///tmp/invalid/svn/repo/hello hello\n" + "file:///tmp/invalid/svn/repo/foo.sh foo.sh" + ) + }, + ), + ], + ) + + 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_with_valid_externals_modification( + swh_storage, repo_url, external_repo_url, tmp_path +): + # first commit on external + add_commit( + external_repo_url, + "Create some directories and files in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/hello/hello-world", + properties={"svn:executable": "*"}, + data=b"#!/bin/bash\necho Hello World !", + ), + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/bar/bar.sh", + properties={"svn:executable": "*"}, + data=b"#!/bin/bash\necho bar", + ), + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="foo.sh", + properties={"svn:executable": "*"}, + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + # first commit + add_commit( + repo_url, + ("Set svn:externals property on trunk/externals path of repository to load."), + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/externals/", + properties={ + "svn:externals": ( + f"{svn_urljoin(external_repo_url, 'code/hello')} src/code/hello\n" # noqa + f"{svn_urljoin(external_repo_url, 'foo.sh')} src/foo.sh\n" + ) + }, + ), + ], + ) + + # second commit + add_commit( + repo_url, + ( + "Modify svn:externals property on trunk/externals path of repository to load." # noqa + ), + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/externals/", + properties={ + "svn:externals": ( + f"{svn_urljoin(external_repo_url, 'code/bar')} src/code/bar\n" # noqa + f"{svn_urljoin(external_repo_url, 'foo.sh')} src/foo.sh\n" + ) + }, + ), + ], + ) + + 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_with_valid_externals_and_versioned_path( + swh_storage, repo_url, external_repo_url, tmp_path +): + # first commit on external + add_commit( + external_repo_url, + "Create a file in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/script.sh", + data=b"#!/bin/bash\necho Hello World !", + ), + ], + ) + + # first commit + add_commit( + repo_url, + "Add file with same name but different content in main repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/script.sh", + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + # second commit + add_commit( + repo_url, + "Add externals targeting the versioned file", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={ + "svn:externals": ( + f"{svn_urljoin(external_repo_url, 'code/script.sh')} script.sh" # noqa + ) + }, + ), + ], + ) + + # third commit + add_commit( + repo_url, + "Modify the versioned file", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/script.sh", + data=b"#!/bin/bash\necho bar", + ), + ], + ) + + 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_with_invalid_externals_and_versioned_path( + swh_storage, repo_url, tmp_path +): + + # first commit + add_commit( + repo_url, + "Add file in main repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/script.sh", + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + # second commit + add_commit( + repo_url, + "Add invalid externals targeting the versioned file", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={ + "svn:externals": ( + "file:///tmp/invalid/svn/repo/code/script.sh script.sh" + ) + }, + ), + ], + ) + + 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_set_externals_then_remove_and_add_as_local( + swh_storage, repo_url, external_repo_url, tmp_path +): + # first commit on external + add_commit( + external_repo_url, + "Create a file in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/script.sh", + data=b"#!/bin/bash\necho Hello World !", + ), + ], + ) + + # first commit + add_commit( + repo_url, + "Add trunk directory and set externals", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={ + "svn:externals": (f"{svn_urljoin(external_repo_url, 'code')} code") + }, + ), + ], + ) + + # second commit + add_commit( + repo_url, + "Unset externals on trunk and add remote path as local path", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={"svn:externals": None}, + ), + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/code/script.sh", + data=b"#!/bin/bash\necho Hello World !", + ), + ], + ) + + 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_set_invalid_externals_then_remove(swh_storage, repo_url, tmp_path): + + # first commit + add_commit( + repo_url, + "Add trunk directory and set invalid external", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={ + "svn:externals": "file:///tmp/invalid/svn/repo/code external/code" + }, + ), + ], + ) + + # second commit + add_commit( + repo_url, + "Unset externals on trunk", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={"svn:externals": None}, + ), + ], + ) + + 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_set_externals_with_versioned_file_overlap( + swh_storage, repo_url, external_repo_url, tmp_path +): + # first commit on external + add_commit( + external_repo_url, + "Create a file in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/script.sh", + data=b"#!/bin/bash\necho Hello World !", + ), + ], + ) + + # first commit + add_commit( + repo_url, + "Add file with same name as in the external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/script.sh", + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + # second commit + add_commit( + repo_url, + "Set external on trunk overlapping versioned file", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={ + "svn:externals": ( + f"{svn_urljoin(external_repo_url, 'code/script.sh')} script.sh" + ) + }, + ), + ], + ) + + # third commit + add_commit( + repo_url, + "Unset externals on trunk", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={"svn:externals": None}, + ), + ], + ) + + 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_dump_loader_relative_externals_detection( + swh_storage, repo_url, external_repo_url, tmp_path +): + + add_commit( + external_repo_url, + "Create a file in external repository.", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="project1/foo.sh", + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + add_commit( + external_repo_url, + "Create another file in repository to load.", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="project2/bar.sh", + data=b"#!/bin/bash\necho bar", + ), + ], + ) + + external_url = f"{external_repo_url.replace('file://', '//')}/project2/bar.sh" + add_commit( + repo_url, + "Set external relative to URL scheme in repository to load", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="project1/", + properties={"svn:externals": (f"{external_url} bar.sh")}, + ), + ], + ) + + loader = SvnLoaderFromRemoteDump( + 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 loader.svnrepo.has_relative_externals + + add_commit( + repo_url, + "Unset external in repository to load", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="project1/", + properties={"svn:externals": None}, + ), + ], + ) + + loader = SvnLoaderFromRemoteDump( + 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 not loader.svnrepo.has_relative_externals + + +def test_loader_externals_cache(swh_storage, repo_url, external_repo_url, tmp_path): + + # first commit on external + add_commit( + external_repo_url, + "Create some directories and files in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/hello/hello-world", + properties={"svn:executable": "*"}, + data=b"#!/bin/bash\necho Hello World !", + ), + ], + ) + + # first commit + add_commit( + repo_url, + "Create repository structure.", + [ + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="project1/",), + CommitChange(change_type=CommitChangeType.AddOrUpdate, path="project2/",), + ], + ) + + external_url = svn_urljoin(external_repo_url, "code/hello") + + # second commit + add_commit( + repo_url, + ( + "Set svn:externals property on trunk/externals path of repository to load." + "One external targets a remote directory and another one a remote file." + ), + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="project1/externals/", + properties={"svn:externals": (f"{external_url} hello\n")}, + ), + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="project2/externals/", + properties={"svn:externals": (f"{external_url} hello\n")}, + ), + ], + ) + + 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 (external_url, None) in loader.svnrepo.swhreplay.editor.externals_cache + + +def test_loader_remove_versioned_path_with_external_overlap( + swh_storage, repo_url, external_repo_url, tmp_path +): + # first commit on external + add_commit( + external_repo_url, + "Create a file in an external repository", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="code/hello.sh", + data=b"#!/bin/bash\necho Hello World !", + ), + ], + ) + + # first commit + add_commit( + repo_url, + "Add a file", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/project/script.sh", + data=b"#!/bin/bash\necho foo", + ), + ], + ) + + # second commit + add_commit( + repo_url, + "Set external on trunk overlapping versioned path", + [ + CommitChange( + change_type=CommitChangeType.AddOrUpdate, + path="trunk/", + properties={ + "svn:externals": ( + f"{svn_urljoin(external_repo_url, 'code')} project/code" + ) + }, + ), + ], + ) + + # third commit + add_commit( + repo_url, + "Remove trunk/project/ versioned path", + [CommitChange(change_type=CommitChangeType.Delete, path="trunk/project/",),], + ) + + 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) diff --git a/swh/loader/svn/tests/test_utils.py b/swh/loader/svn/tests/test_utils.py index 17a3199..d4a91ce 100644 --- a/swh/loader/svn/tests/test_utils.py +++ b/swh/loader/svn/tests/test_utils.py @@ -1,141 +1,383 @@ -# Copyright (C) 2016-2021 The Software Heritage developers +# 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 logging import os import pty import shutil from subprocess import Popen +import pytest + from swh.loader.svn import utils -from swh.model.model import Timestamp def test_outputstream(): stdout_r, stdout_w = pty.openpty() echo = Popen(["echo", "-e", "foo\nbar\nbaz"], stdout=stdout_w) os.close(stdout_w) stdout_stream = utils.OutputStream(stdout_r) lines = [] while True: current_lines, readable = stdout_stream.read_lines() lines += current_lines if not readable: break echo.wait() os.close(stdout_r) assert lines == ["foo", "bar", "baz"] -def test_strdate_to_timestamp(): - """Formatted string date should be converted in timestamp.""" - actual_ts = utils.strdate_to_timestamp("2011-05-31T06:04:39.800722Z") - assert actual_ts == Timestamp(seconds=1306821879, microseconds=800722) - - actual_ts = utils.strdate_to_timestamp("2011-05-31T06:03:39.123450Z") - assert actual_ts == Timestamp(seconds=1306821819, microseconds=123450) - - -def test_strdate_to_timestamp_empty_does_not_break(): - """Empty or None date should be timestamp 0.""" - default_ts = Timestamp(seconds=0, microseconds=0) - assert default_ts == utils.strdate_to_timestamp("") - assert default_ts == utils.strdate_to_timestamp(None) - - def test_init_svn_repo_from_dump(datadir, tmp_path): """Mounting svn repository out of a dump is ok""" dump_name = "penguinsdbtools2018.dump.gz" dump_path = os.path.join(datadir, dump_name) tmp_repo, repo_path = utils.init_svn_repo_from_dump( dump_path, gzip=True, cleanup_dump=False, root_dir=tmp_path ) assert os.path.exists(dump_path), "Dump path should still exists" assert os.path.exists(repo_path), "Repository should exists" def test_init_svn_repo_from_dump_and_cleanup(datadir, tmp_path): """Mounting svn repository with a dump cleanup after is ok""" dump_name = "penguinsdbtools2018.dump.gz" dump_ori_path = os.path.join(datadir, dump_name) dump_path = os.path.join(tmp_path, dump_name) shutil.copyfile(dump_ori_path, dump_path) assert os.path.exists(dump_path) assert os.path.exists(dump_ori_path) tmp_repo, repo_path = utils.init_svn_repo_from_dump( dump_path, gzip=True, root_dir=tmp_path ) assert not os.path.exists(dump_path), "Dump path should no longer exists" assert os.path.exists(repo_path), "Repository should exists" assert os.path.exists(dump_ori_path), "Original dump path should still exists" def test_init_svn_repo_from_dump_and_cleanup_already_done( datadir, tmp_path, mocker, caplog ): """Mounting svn repository out of a dump is ok""" caplog.set_level(logging.INFO, "swh.loader.svn.utils") dump_name = "penguinsdbtools2018.dump.gz" dump_ori_path = os.path.join(datadir, dump_name) mock_remove = mocker.patch("os.remove") mock_remove.side_effect = FileNotFoundError dump_path = os.path.join(tmp_path, dump_name) shutil.copyfile(dump_ori_path, dump_path) assert os.path.exists(dump_path) assert os.path.exists(dump_ori_path) tmp_repo, repo_path = utils.init_svn_repo_from_dump( dump_path, gzip=True, root_dir=tmp_path ) assert os.path.exists(repo_path), "Repository should exists" assert os.path.exists(dump_ori_path), "Original dump path should still exists" assert len(caplog.record_tuples) == 1 assert "Failure to remove" in caplog.record_tuples[0][2] assert mock_remove.called def test_init_svn_repo_from_archive_dump(datadir, tmp_path): """Mounting svn repository out of an archive dump is ok""" dump_name = "penguinsdbtools2018.dump.gz" dump_path = os.path.join(datadir, dump_name) tmp_repo, repo_path = utils.init_svn_repo_from_archive_dump( dump_path, cleanup_dump=False, root_dir=tmp_path ) assert os.path.exists(dump_path), "Dump path should still exists" assert os.path.exists(repo_path), "Repository should exists" def test_init_svn_repo_from_archive_dump_and_cleanup(datadir, tmp_path): """Mounting svn repository out of a dump is ok""" dump_name = "penguinsdbtools2018.dump.gz" dump_ori_path = os.path.join(datadir, dump_name) dump_path = os.path.join(tmp_path, dump_name) shutil.copyfile(dump_ori_path, dump_path) assert os.path.exists(dump_path) assert os.path.exists(dump_ori_path) tmp_repo, repo_path = utils.init_svn_repo_from_archive_dump( dump_path, root_dir=tmp_path ) assert not os.path.exists(dump_path), "Dump path should no longer exists" assert os.path.exists(repo_path), "Repository should exists" assert os.path.exists(dump_ori_path), "Original dump path should still exists" + + +@pytest.mark.parametrize( + "base_url, paths_to_join, expected_result", + [ + ( + "https://svn.example.org", + ["repos", "test"], + "https://svn.example.org/repos/test", + ), + ( + "https://svn.example.org/", + ["repos", "test"], + "https://svn.example.org/repos/test", + ), + ( + "https://svn.example.org/foo", + ["repos", "test"], + "https://svn.example.org/foo/repos/test", + ), + ( + "https://svn.example.org/foo/", + ["/repos", "test/"], + "https://svn.example.org/foo/repos/test", + ), + ("https://svn.example.org/foo", ["../bar"], "https://svn.example.org/bar",), + ], +) +def test_svn_urljoin(base_url, paths_to_join, expected_result): + assert utils.svn_urljoin(base_url, *paths_to_join) == expected_result + + +@pytest.mark.parametrize( + "external, dir_path, repo_url, expected_result", + [ + # subversion < 1.5 + ( + "third-party/sounds http://svn.example.com/repos/sounds", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/sounds", "http://svn.example.com/repos/sounds", None, False), + ), + ( + "third-party/skins -r148 http://svn.example.com/skinproj", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/skins", "http://svn.example.com/skinproj", 148, False), + ), + ( + "third-party/skins/toolkit -r21 http://svn.example.com/skin-maker", + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/skins/toolkit", + "http://svn.example.com/skin-maker", + 21, + False, + ), + ), + # subversion >= 1.5 + ( + " http://svn.example.com/repos/sounds third-party/sounds", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/sounds", "http://svn.example.com/repos/sounds", None, False), + ), + ( + "-r148 http://svn.example.com/skinproj third-party/skins", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/skins", "http://svn.example.com/skinproj", 148, False), + ), + ( + "-r 21 http://svn.example.com/skin-maker third-party/skins/toolkit", + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/skins/toolkit", + "http://svn.example.com/skin-maker", + 21, + False, + ), + ), + ( + "http://svn.example.com/repos/sounds third-party/sounds", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/sounds", "http://svn.example.com/repos/sounds", None, False), + ), + ( + "http://svn.example.com/skinproj@148 third-party/skins", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/skins", "http://svn.example.com/skinproj", 148, False), + ), + ( + "http://anon:anon@svn.example.com/skin-maker@21 third-party/skins/toolkit", + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/skins/toolkit", + "http://anon:anon@svn.example.com/skin-maker", + 21, + False, + ), + ), + ( + "-r21 http://anon:anon@svn.example.com/skin-maker third-party/skins/toolkit", # noqa + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/skins/toolkit", + "http://anon:anon@svn.example.com/skin-maker", + 21, + False, + ), + ), + ( + "-r21 http://anon:anon@svn.example.com/skin-maker@21 third-party/skins/toolkit", # noqa + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/skins/toolkit", + "http://anon:anon@svn.example.com/skin-maker", + 21, + False, + ), + ), + # subversion >= 1.5, relative external definitions + ( + "^/sounds third-party/sounds", + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/sounds", + "http://svn.example.org/repos/test/sounds", + None, + False, + ), + ), + ( + "/skinproj@148 third-party/skins", + "trunk/externals", + "http://svn.example.org/repos/test", + ("third-party/skins", "http://svn.example.org/skinproj", 148, True), + ), + ( + "//svn.example.com/skin-maker@21 third-party/skins/toolkit", + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "third-party/skins/toolkit", + "http://svn.example.com/skin-maker", + 21, + True, + ), + ), + ( + "../skins skins", + "trunk/externals", + "http://svn.example.org/repos/test", + ("skins", "http://svn.example.org/repos/test/trunk/skins", None, False), + ), + ( + "../skins skins", + "trunk/externals", + "http://svn.example.org/repos/test", + ("skins", "http://svn.example.org/repos/test/trunk/skins", None, False), + ), + # subversion >= 1.6 + ( + 'http://svn.thirdparty.com/repos/My%20Project "My Project"', + "trunk/externals", + "http://svn.example.org/repos/test", + ("My Project", "http://svn.thirdparty.com/repos/My%20Project", None, False), + ), + ( + 'http://svn.thirdparty.com/repos/My%20%20%20Project "My Project"', + "trunk/externals", + "http://svn.example.org/repos/test", + ( + "My Project", + "http://svn.thirdparty.com/repos/My%20%20%20Project", + None, + False, + ), + ), + ( + 'http://svn.thirdparty.com/repos/%22Quotes%20Too%22 \\"Quotes\\ Too\\"', + "trunk/externals", + "http://svn.example.org/repos/test", + ( + '"Quotes Too"', + "http://svn.thirdparty.com/repos/%22Quotes%20Too%22", + None, + False, + ), + ), + ( + 'http://svn.thirdparty.com/repos/%22Quotes%20%20%20Too%22 \\"Quotes\\ \\ \\ Too\\"', # noqa + "trunk/externals", + "http://svn.example.org/repos/test", + ( + '"Quotes Too"', + "http://svn.thirdparty.com/repos/%22Quotes%20%20%20Too%22", + None, + False, + ), + ), + # edge cases + ( + '-r1 http://svn.thirdparty.com/repos/test "trunk/PluginFramework"', + "trunk/externals", + "http://svn.example.org/repos/test", + ("trunk/PluginFramework", "http://svn.thirdparty.com/repos/test", 1, False), + ), + ( + "external -r 9 http://svn.thirdparty.com/repos/test", + "tags", + "http://svn.example.org/repos/test", + ("external", "http://svn.thirdparty.com/repos/test", 9, False), + ), + ( + "./external http://svn.thirdparty.com/repos/test", + "tags", + "http://svn.example.org/repos/test", + ("external", "http://svn.thirdparty.com/repos/test", None, False), + ), + ( + "external ttp://svn.thirdparty.com/repos/test", + "tags", + "http://svn.example.org/repos/test", + ("external", "ttp://svn.thirdparty.com/repos/test", None, False), + ), + ( + "C:\\code\\repo\\external http://svn.thirdparty.com/repos/test", + "tags", + "http://svn.example.org/repos/test", + ("C:coderepoexternal", "http://svn.thirdparty.com/repos/test", None, False), + ), + ( + "C:\\\\code\\\\repo\\\\external http://svn.thirdparty.com/repos/test", + "tags", + "http://svn.example.org/repos/test", + ( + "C:\\code\\repo\\external", + "http://svn.thirdparty.com/repos/test", + None, + False, + ), + ), + ], +) +def test_parse_external_definition(external, dir_path, repo_url, expected_result): + assert ( + utils.parse_external_definition(external, dir_path, repo_url) == expected_result + ) diff --git a/swh/loader/svn/utils.py b/swh/loader/svn/utils.py index 03927a2..5637346 100644 --- a/swh/loader/svn/utils.py +++ b/swh/loader/svn/utils.py @@ -1,195 +1,294 @@ -# Copyright (C) 2016-2021 The Software Heritage developers +# 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 errno import logging import os +import re import shutil from subprocess import PIPE, Popen, call import tempfile -from typing import Tuple - -from dateutil import parser - -from swh.model.model import Optional, Timestamp +from typing import Optional, Tuple +from urllib.parse import urlparse logger = logging.getLogger(__name__) -def strdate_to_timestamp(strdate: Optional[str]) -> Timestamp: - """Convert a string date to an int timestamp. - - Args: - strdate: A string representing a date with format like - 'YYYY-mm-DDTHH:MM:SS.800722Z' - - Returns: - A couple of integers: seconds, microseconds - - """ - if strdate: - # TODO: Migrate to iso8601 if possible - dt = parser.parse(strdate) - ts = { - "seconds": int(dt.timestamp()), - "microseconds": dt.microsecond, - } - else: # epoch - ts = {"seconds": 0, "microseconds": 0} - return Timestamp.from_dict(ts) - - class OutputStream: """Helper class to read lines from a program output while it is running Args: fileno (int): File descriptor of a program output stream opened in text mode """ def __init__(self, fileno): self._fileno = fileno self._buffer = "" def read_lines(self): """ Read available lines from the output stream and return them. Returns: Tuple[List[str], bool]: A tuple whose first member is the read lines and second member a boolean indicating if there are still some other lines available to read. """ try: output = os.read(self._fileno, 1000).decode() except OSError as e: if e.errno != errno.EIO: raise output = "" output = output.replace("\r\n", "\n") lines = output.split("\n") lines[0] = self._buffer + lines[0] if output: self._buffer = lines[-1] return (lines[:-1], True) else: self._buffer = "" if len(lines) == 1 and not lines[0]: lines = [] return (lines, False) def init_svn_repo_from_dump( dump_path: str, prefix: Optional[str] = None, suffix: Optional[str] = None, root_dir: str = "/tmp", gzip: bool = False, cleanup_dump: bool = True, ) -> Tuple[str, str]: """Given a path to a svn dump, initialize an svn repository with the content of said dump. Args: dump_path: The dump to the path prefix: optional prefix file name for the working directory suffix: optional suffix file name for the working directory root_dir: the root directory where the working directory is created gzip: Boolean to determine whether we treat the dump as compressed or not. cleanup_dump: Whether we want this function call to clean up the dump at the end of the repository initialization. Raises: ValueError in case of failure to run the command to uncompress and load the dump. Returns: A tuple: - temporary folder: containing the mounted repository - repo_path: path to the mounted repository inside the temporary folder """ project_name = os.path.basename(os.path.dirname(dump_path)) temp_dir = tempfile.mkdtemp(prefix=prefix, suffix=suffix, dir=root_dir) try: repo_path = os.path.join(temp_dir, project_name) # create the repository that will be loaded with the dump cmd = ["svnadmin", "create", repo_path] r = call(cmd) if r != 0: raise ValueError( "Failed to initialize empty svn repo for %s" % project_name ) read_dump_cmd = ["cat", dump_path] if gzip: read_dump_cmd = ["gzip", "-dc", dump_path] with Popen(read_dump_cmd, stdout=PIPE) as dump: # load dump and bypass properties validation as Unicode decoding errors # are already handled in loader implementation (see _ra_codecs_error_handler # in ra.py) cmd = ["svnadmin", "load", "-q", "--bypass-prop-validation", repo_path] r = call(cmd, stdin=dump.stdout) if r != 0: raise ValueError( "Failed to mount the svn dump for project %s" % project_name ) return temp_dir, repo_path except Exception as e: shutil.rmtree(temp_dir) raise e finally: if cleanup_dump: try: # At this time, the temporary svn repository is mounted from the dump or # the svn repository failed to mount. Either way, we can drop the dump. os.remove(dump_path) assert not os.path.exists(dump_path) except OSError as e: logger.warn("Failure to remove the dump %s: %s", dump_path, e) def init_svn_repo_from_archive_dump( archive_path: str, prefix: Optional[str] = None, suffix: Optional[str] = None, root_dir: str = "/tmp", cleanup_dump: bool = True, ) -> Tuple[str, str]: """Given a path to an archive containing an svn dump, initializes an svn repository with the content of the uncompressed dump. Args: archive_path: The archive svn dump path prefix: optional prefix file name for the working directory suffix: optional suffix file name for the working directory root_dir: the root directory where the working directory is created gzip: Boolean to determine whether we treat the dump as compressed or not. cleanup_dump: Whether we want this function call to clean up the dump at the end of the repository initialization. Raises: ValueError in case of failure to run the command to uncompress and load the dump. Returns: A tuple: - temporary folder: containing the mounted repository - repo_path: path to the mounted repository inside the temporary folder """ return init_svn_repo_from_dump( archive_path, prefix=prefix, suffix=suffix, root_dir=root_dir, gzip=True, cleanup_dump=cleanup_dump, ) + + +def svn_urljoin(base_url: str, *args) -> str: + """Join a base URL and a list of paths in a SVN way. + + For instance: + + - svn_urljoin("http://example.org", "foo", "bar") + will return "https://example.org/foo/bar + + - svn_urljoin("http://example.org/foo", "../bar") + will return "https://example.org/bar + + Args: + base_url: Base URL to join paths with + args: path components + + Returns: + The joined URL + + """ + parsed_url = urlparse(base_url) + path = os.path.abspath( + os.path.join(parsed_url.path or "/", *[arg.strip("/") for arg in args]) + ) + return f"{parsed_url.scheme}://{parsed_url.netloc}{path}" + + +def parse_external_definition( + external: str, dir_path: str, repo_url: str +) -> Tuple[str, str, Optional[int], bool]: + """Parse a subversion external definition. + + Args: + external: an external definition, extracted from the lines split of a + svn:externals property value + dir_path: The path of the directory in the subversion repository where + the svn:externals property was set + repo_url: URL of the subversion repository + + Returns: + A tuple with the following members: + + - path relative to dir_path where the external should be exported + - URL of the external to export + - optional revision of the external to export + - boolean indicating if the external URL is relative to the repository + URL and targets a path not in the repository + + """ + path = "" + external_url = "" + revision = None + relative_url = False + prev_part = None + # turn multiple spaces into a single one and split on space + for external_part in external.split(): + if prev_part == "-r": + # parse revision in the form "-r XXX" + revision = int(external_part) + elif external_part.startswith("-r") and external_part != "-r": + # parse revision in the form "-rXXX" + revision = int(external_part[2:]) + elif external_part.startswith("^/"): + # URL relative to the root of the repository in which the svn:externals + # property is versioned + external_url = svn_urljoin(repo_url, external_part[2:]) + elif external_part.startswith("//"): + # URL relative to the scheme of the URL of the directory on which the + # svn:externals property is set + scheme = urlparse(repo_url).scheme + external_url = f"{scheme}:{external_part}" + relative_url = not external_url.startswith(repo_url) + elif external_part.startswith("/"): + # URL relative to the root URL of the server on which the svn:externals + # property is versioned + parsed_url = urlparse(repo_url) + root_url = f"{parsed_url.scheme}://{parsed_url.netloc}" + external_url = svn_urljoin(root_url, external_part) + relative_url = not external_url.startswith(repo_url) + elif external_part.startswith("../"): + # URL relative to the URL of the directory on which the svn:externals + # property is set + external_url = svn_urljoin(repo_url, dir_path, external_part) + relative_url = not external_url.startswith(repo_url) + elif re.match(r"^.*://.*", external_part): + # absolute external URL + external_url = external_part + # subversion >= 1.6 added a quoting and escape mechanism to the syntax so + # that the path of the external working copy may contain whitespace. + elif external_part.startswith('\\"'): + external_split = external.split('\\"') + path = [ + e.replace("\\ ", " ") + for e in external_split + if e.startswith(external_part[2:]) + ][0] + path = f'"{path}"' + elif external_part.endswith('\\"'): + continue + elif external_part.startswith('"'): + external_split = external.split('"') + path_prefix = external_part.strip('"') + path = next(iter([e for e in external_split if e.startswith(path_prefix)])) + elif external_part.endswith('"'): + continue + elif not external_part.startswith("\\") and external_part != "-r": + # path of the external relative to dir_path + path = external_part.replace("\\\\", "\\") + if path == external_part: + path = external_part.replace("\\", "") + path = path.lstrip("./") + prev_part = external_part + if "@" in external_url: + # try to extract revision number if external URL is in the form + # http://svn.example.org/repos/test/path@XXX + url, revision_s = external_url.rsplit("@", maxsplit=1) + try: + revision = int(revision_s) + external_url = url + except ValueError: + # handle URL like http://user@svn.example.org/ + pass + return (path, external_url, revision, relative_url) diff --git a/tox.ini b/tox.ini index fc2d4d5..47d1062 100644 --- a/tox.ini +++ b/tox.ini @@ -1,74 +1,74 @@ [tox] envlist=black,flake8,mypy,py3 [testenv] extras = testing deps = pytest-cov swh.scheduler[testing] >= 0.5.0 dev: pdbpp commands = pytest --cov={envsitepackagesdir}/swh/loader/svn \ {envsitepackagesdir}/swh/loader/svn \ --cov-branch {posargs} [testenv:black] skip_install = true deps = black==19.10b0 commands = {envpython} -m black --check swh [testenv:flake8] skip_install = true deps = flake8 commands = {envpython} -m flake8 [testenv:mypy] extras = testing deps = - mypy + mypy==0.920 commands = mypy swh # build documentation outside swh-environment using the current # git HEAD of swh-docs, is executed on CI for each diff to prevent # breaking doc build [testenv:sphinx] whitelist_externals = make usedevelop = true extras = testing deps = # fetch and install swh-docs in develop mode -e git+https://forge.softwareheritage.org/source/swh-docs#egg=swh.docs setenv = SWH_PACKAGE_DOC_TOX_BUILD = 1 # turn warnings into errors SPHINXOPTS = -W commands = make -I ../.tox/sphinx/src/swh-docs/swh/ -C docs # build documentation only inside swh-environment using local state # of swh-docs package [testenv:sphinx-dev] whitelist_externals = make usedevelop = true extras = testing deps = # install swh-docs in develop mode -e ../swh-docs setenv = SWH_PACKAGE_DOC_TOX_BUILD = 1 # turn warnings into errors SPHINXOPTS = -W commands = make -I ../.tox/sphinx-dev/src/swh-docs/swh/ -C docs