From 354bfaf28e031cc6731d1ed3f220a4311f4f52ef Mon Sep 17 00:00:00 2001 From: taoky Date: Wed, 30 Oct 2024 20:30:39 +0800 Subject: [PATCH] Add SHADOWMIRE_IGNORE_THRESHOLD option Related: https://github.com/tuna/issues/issues/2061 --- shadowmire.py | 40 +++++++++++++++++++++++++++++++++++----- 1 file changed, 35 insertions(+), 5 deletions(-) diff --git a/shadowmire.py b/shadowmire.py index 843cb9b..eab7fbb 100755 --- a/shadowmire.py +++ b/shadowmire.py @@ -42,6 +42,8 @@ WORKERS = int(os.environ.get("SHADOWMIRE_WORKERS", "3")) IOWORKERS = int(os.environ.get("SHADOWMIRE_IOWORKERS", "2")) # A safety net -- to avoid upstream issues casuing too many packages removed when determinating sync plan. MAX_DELETION = int(os.environ.get("SHADOWMIRE_MAX_DELETION", "50000")) +# Sometimes PyPI is not consistent -- new packages could not be fetched. This option tries to avoid permanently mark that kind of package as nonexist. +IGNORE_THRESHOLD = int(os.environ.get("SHADOWMIRE_IGNORE_THRESHOLD", "1024")) # https://github.com/pypa/bandersnatch/blob/a05af547f8d1958217ef0dc0028890b1839e6116/src/bandersnatch_filter_plugins/prerelease_name.py#L18C1-L23C6 PRERELEASE_PATTERNS = ( @@ -337,6 +339,9 @@ class PyPI: del ret[key] return ret + def changelog_last_serial(self) -> int: + return self.xmlrpc_client.changelog_last_serial() # type: ignore + def get_package_metadata(self, package_name: str) -> dict: req = self.session.get(urljoin(self.host, f"pypi/{package_name}/json")) if req.status_code == 404: @@ -828,15 +833,18 @@ class SyncPyPI(SyncBase): ) -> None: self.pypi = PyPI() self.session = create_requests_session() + self.last_serial: Optional[int] = None + self.remote_packages: Optional[dict[str, int]] = None super().__init__(basedir, local_db, sync_packages) def fetch_remote_versions(self) -> dict[str, int]: - ret = self.pypi.list_packages_with_serial() - logger.info("Remote has %s packages", len(ret)) + self.last_serial = self.pypi.changelog_last_serial() + self.remote_packages = self.pypi.list_packages_with_serial() + logger.info("Remote has %s packages", len(self.remote_packages)) with overwrite(self.basedir / "remote.json") as f: - json.dump(ret, f) + json.dump(self.remote_packages, f) logger.info("File saved to remote.json.") - return ret + return self.remote_packages def do_update( self, @@ -852,9 +860,31 @@ class SyncPyPI(SyncBase): meta_original = deepcopy(meta) logger.debug("%s meta: %s", package_name, meta) except PackageNotFoundError: + if ( + self.remote_packages is not None + and package_name in self.remote_packages + ): + recorded_serial = self.remote_packages[package_name] + else: + recorded_serial = None + if ( + recorded_serial is not None + and self.last_serial is not None + and abs(recorded_serial - self.last_serial) < IGNORE_THRESHOLD + ): + logger.warning( + "%s missing from upstream (its serial %s, remote last serial %s), try next time...", + package_name, + recorded_serial, + self.last_serial, + ) + return None + logger.warning( - "%s missing from upstream, remove and ignore in the future.", + "%s missing from upstream (its serial %s, remote last serial %s), remove and ignore in the future.", package_name, + recorded_serial, + self.last_serial, ) # try remove it locally, if it does not exist upstream self.do_remove(package_name, use_db=False)