Source code for datalad.tests.utils

# emacs: -*- mode: python; py-indent-offset: 4; tab-width: 4; indent-tabs-mode: nil; coding: utf-8 -*-
# ex: set sts=4 ts=4 sw=4 et:
# ## ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ##
#
#   See COPYING file distributed along with the datalad package for the
#   copyright and license terms.
#
# ## ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ##
"""Miscellaneous utilities to assist with testing"""

import base64
import glob
import gzip
import inspect
import logging
import lzma
import multiprocessing
import multiprocessing.queues
import os
import platform
import random
import re
import shutil
import socket
import ssl
import stat
import tempfile
import textwrap
import time
import warnings
from contextlib import contextmanager
from difflib import unified_diff
from fnmatch import fnmatch
from functools import wraps
from http.server import (
    HTTPServer,
    SimpleHTTPRequestHandler,
)
from json import dumps
from os.path import (
    curdir,
    exists,
)
from os.path import join as opj
from os.path import relpath
from os.path import split as pathsplit
from unittest.mock import patch

from nose import SkipTest
from nose.plugins.attrib import attr
from nose.tools import (
    assert_equal,
    assert_false,
    assert_greater,
    assert_greater_equal,
)
from nose.tools import assert_in
from nose.tools import assert_in as in_
from nose.tools import (
    assert_is,
    assert_is_instance,
    assert_is_none,
    assert_is_not,
    assert_is_not_none,
    assert_not_equal,
    assert_not_in,
    assert_not_is_instance,
    assert_raises,
    assert_set_equal,
    assert_true,
    eq_,
    make_decorator,
    ok_,
    raises,
)

import datalad.utils as ut
from datalad import cfg as dl_cfg
from datalad import (
    conftest,
    ssh_manager,
)
from datalad.cmd import (
    GitWitlessRunner,
    KillOutput,
    StdOutErrCapture,
    WitlessRunner,
)
from datalad.core.local.repo import repo_from_path
from datalad.utils import (
    Path,
    ensure_unicode,
    get_encoding_info,
    get_envvars_info,
    get_home_envvars,
)

from .. import utils
from ..consts import ARCHIVES_TEMP_DIR
from ..dochelpers import borrowkwargs
from ..support.exceptions import (
    CommandError,
    CommandNotAvailableError,
)
from ..support.external_versions import external_versions
from ..support.keyring_ import MemoryKeyring
from ..support.network import RI
from ..support.vcr_ import *
# TODO this must go
from ..utils import *
from . import _TEMP_PATHS_GENERATED

# temp paths used by clones
_TEMP_PATHS_CLONES = set()


# Additional indicators
on_travis = bool(os.environ.get('TRAVIS', False))

if external_versions["cmd:git"] >= "2.28":
    # The specific value here doesn't matter, but it should not be the default
    # from any Git version to test that we work with custom values.
    DEFAULT_BRANCH = "dl-test-branch"  # Set by setup_package().
else:
    DEFAULT_BRANCH = "master"

if external_versions["cmd:git"] >= "2.30.0":
    # The specific value here doesn't matter, but it should not be the default
    # from any Git version to test that we work with custom values.
    DEFAULT_REMOTE = "dl-test-remote"  # Set by setup_package().
else:
    DEFAULT_REMOTE = "origin"

# additional shortcuts
neq_ = assert_not_equal
nok_ = assert_false

lgr = logging.getLogger("datalad.tests.utils")

# To store settings which setup_package changes and teardown_package should return
_test_states = {
    'loglevel': None,
    'env': {},
}

[docs]def setup_package(): import os import tempfile from io import StringIO as OrigStringIO from pathlib import Path from nose.ext import dtcompat from nose.plugins import ( capture, multiprocess, plugintest, ) from datalad import consts, lgr from datalad.support.annexrepo import AnnexRepo from datalad.support.external_versions import external_versions from datalad.tests.utils import ( DEFAULT_BRANCH, DEFAULT_REMOTE, HTTPPath, ) from datalad.ui import ui from datalad.utils import ( make_tempfile, on_osx, ) if on_osx: # enforce honoring TMPDIR (see gh-5307) tempfile.tempdir = os.environ.get('TMPDIR', tempfile.gettempdir()) _test_states['env'] = {} def set_envvar(v, val): """Memoize and then set env var""" _test_states['env'][v] = os.environ.get(v, None) os.environ[v] = val _test_states['DATASETS_TOPURL'] = consts.DATASETS_TOPURL consts.DATASETS_TOPURL = 'https://datasets-tests.datalad.org/' set_envvar('DATALAD_DATASETS_TOPURL', consts.DATASETS_TOPURL) set_envvar("GIT_CONFIG_PARAMETERS", "'init.defaultBranch={}' 'clone.defaultRemoteName={}'" .format(DEFAULT_BRANCH, DEFAULT_REMOTE)) def prep_tmphome(): # re core.askPass: # Don't let git ask for credentials in CI runs. Note, that this variable # technically is not a flag, but an executable (which is why name and value # are a bit confusing here - we just want a no-op basically). The environment # variable GIT_ASKPASS overwrites this, but neither env var nor this config # are supported by git-credential on all systems and git versions (most recent # ones should work either way, though). Hence use both across CI builds. gitconfig = """\ [user] name = DataLad Tester email = test@example.com [core] askPass = [datalad "log"] exc = 1 [annex "security"] # from annex 6.20180626 file:/// and http://localhost access isn't # allowed by default allowed-url-schemes = http https file allowed-http-addresses = all """ # TODO: split into a function + context manager with make_tempfile(mkdir=True) as new_home: pass # register for clean-up on exit _TEMP_PATHS_GENERATED.append(new_home) # populate default config new_home = Path(new_home) new_home.mkdir(parents=True, exist_ok=True) cfg_file = new_home / '.gitconfig' cfg_file.write_text(gitconfig) return new_home, cfg_file if external_versions['cmd:git'] < "2.32": # To overcome pybuild overriding HOME but us possibly wanting our # own HOME where we pre-setup git for testing (name, email) if 'GIT_HOME' in os.environ: set_envvar('HOME', os.environ['GIT_HOME']) else: # we setup our own new HOME, the BEST and HUGE one new_home, _ = prep_tmphome() for v, val in get_home_envvars(new_home).items(): set_envvar(v, val) else: _, cfg_file = prep_tmphome() set_envvar('GIT_CONFIG_GLOBAL', str(cfg_file)) # Re-load ConfigManager, since otherwise it won't consider global config # from new $HOME (see gh-4153 dl_cfg.reload(force=True) # datalad.locations.sockets has likely changed. Discard any cached values. ssh_manager._socket_dir = None # To overcome pybuild by default defining http{,s}_proxy we would need # to define them to e.g. empty value so it wouldn't bother touching them. # But then haskell libraries do not digest empty value nicely, so we just # pop them out from the environment for ev in ('http_proxy', 'https_proxy'): if ev in os.environ and not (os.environ[ev]): lgr.debug("Removing %s from the environment since it is empty", ev) os.environ.pop(ev) DATALAD_LOG_LEVEL = os.environ.get('DATALAD_LOG_LEVEL', None) if DATALAD_LOG_LEVEL is None: # very very silent. Tests introspecting logs should use # swallow_logs(new_level=...) _test_states['loglevel'] = lgr.getEffectiveLevel() lgr.setLevel(100) # And we should also set it within environ so underlying commands also stay silent set_envvar('DATALAD_LOG_LEVEL', '100') else: # We are not overriding them, since explicitly were asked to have some log level _test_states['loglevel'] = None # Prevent interactive credential entry (note "true" is the command to run) # See also the core.askPass setting above set_envvar('GIT_ASKPASS', 'true') # Set to non-interactive UI _test_states['ui_backend'] = ui.backend # obtain() since that one consults for the default value ui.set_backend(dl_cfg.obtain('datalad.tests.ui.backend')) # Monkey patch nose so it does not ERROR out whenever code asks for fileno # of the output. See https://github.com/nose-devs/nose/issues/6 class StringIO(OrigStringIO): fileno = lambda self: 1 encoding = None dtcompat.StringIO = StringIO capture.StringIO = StringIO multiprocess.StringIO = StringIO plugintest.StringIO = StringIO # in order to avoid having to fiddle with rather uncommon # file:// URLs in the tests, have a standard HTTP server # that serves an 'httpserve' directory in the test HOME # the URL will be available from datalad.test_http_server.url # Start the server only if not running already # Relevant: we have test_misc.py:test_test which runs datalad.test but # not doing teardown, so the original server might never get stopped if conftest.test_http_server is None: serve_path = tempfile.mkdtemp( dir=dl_cfg.get("datalad.tests.temp.dir"), prefix='httpserve', ) conftest.test_http_server = HTTPPath(serve_path) conftest.test_http_server.start() _TEMP_PATHS_GENERATED.append(serve_path) if dl_cfg.obtain('datalad.tests.setup.testrepos'): lgr.debug("Pre-populating testrepos") from datalad.tests.utils import with_testrepos with_testrepos()(lambda repo: 1)()
[docs]def teardown_package(): import os from datalad import consts from datalad.support.annexrepo import AnnexRepo from datalad.support.cookies import cookies_db from datalad.support.external_versions import external_versions as ev from datalad.tests.utils import ( OBSCURE_FILENAME, rmtemp, ) from datalad.ui import ui lgr.debug("Printing versioning information collected so far") # Query for version of datalad, so it is included in ev.dumps below - useful while # testing extensions where version of datalad might differ in the environment. ev['datalad'] print(ev.dumps(query=True)) try: print("Obscure filename: str=%s repr=%r" % (OBSCURE_FILENAME.encode('utf-8'), OBSCURE_FILENAME)) except UnicodeEncodeError as exc: ce = CapturedException(exc) print("Obscure filename failed to print: %s" % ce) def print_dict(d): return " ".join("%s=%r" % v for v in d.items()) print("Encodings: %s" % print_dict(get_encoding_info())) print("Environment: %s" % print_dict(get_envvars_info())) if os.environ.get('DATALAD_TESTS_NOTEARDOWN'): return ui.set_backend(_test_states['ui_backend']) if _test_states['loglevel'] is not None: lgr.setLevel(_test_states['loglevel']) if conftest.test_http_server: conftest.test_http_server.stop() conftest.test_http_server = None else: lgr.debug("For some reason global http_server was not set/running, thus not stopping") if len(_TEMP_PATHS_GENERATED): msg = "Removing %d dirs/files: %s" % (len(_TEMP_PATHS_GENERATED), ', '.join(_TEMP_PATHS_GENERATED)) else: msg = "Nothing to remove" lgr.debug("Teardown tests. " + msg) for path in _TEMP_PATHS_GENERATED: rmtemp(str(path), ignore_errors=True) # restore all the env variables for v, val in _test_states['env'].items(): if val is not None: os.environ[v] = val else: os.environ.pop(v) # Re-establish correct global config after changing $HOME. # Might be superfluous, since after teardown datalad.cfg shouldn't be # needed. However, maintaining a consistent state seems a good thing # either way. dl_cfg.reload(force=True) ssh_manager._socket_dir = None consts.DATASETS_TOPURL = _test_states['DATASETS_TOPURL'] cookies_db.close() AnnexRepo._ALLOW_LOCAL_URLS = False # stay safe!
[docs]def skip_if_no_module(module): try: imp = __import__(module) except Exception as exc: raise SkipTest("Module %s fails to load" % module) from exc
[docs]def skip_if_scrapy_without_selector(): """A little helper to skip some tests which require recent scrapy""" try: import scrapy from scrapy.selector import Selector except ImportError: from nose import SkipTest raise SkipTest( "scrapy misses Selector (too old? version: %s)" % getattr(scrapy, '__version__'))
[docs]def skip_if_url_is_not_available(url, regex=None): # verify that dataset is available from datalad.downloaders.base import DownloadError from datalad.downloaders.providers import Providers providers = Providers.from_config_files() try: content = providers.fetch(url) if regex and re.search(regex, content): raise SkipTest("%s matched %r -- skipping the test" % (url, regex)) except DownloadError: raise SkipTest("%s failed to download" % url)
[docs]def check_not_generatorfunction(func): """Internal helper to verify that we are not decorating generator tests""" if inspect.isgeneratorfunction(func): raise RuntimeError("{}: must not be decorated, is a generator test" .format(func.__name__))
[docs]def skip_if_no_network(func=None): """Skip test completely in NONETWORK settings If not used as a decorator, and just a function, could be used at the module level """ check_not_generatorfunction(func) def check_and_raise(): if dl_cfg.get('datalad.tests.nonetwork'): raise SkipTest("Skipping since no network settings") if func: @wraps(func) @attr('network') @attr('skip_if_no_network') def _wrap_skip_if_no_network(*args, **kwargs): check_and_raise() return func(*args, **kwargs) return _wrap_skip_if_no_network else: check_and_raise()
[docs]def skip_if_on_windows(func=None): """Skip test completely under Windows """ check_not_generatorfunction(func) def check_and_raise(): if on_windows: raise SkipTest("Skipping on Windows") if func: @wraps(func) @attr('skip_if_on_windows') def _wrap_skip_if_on_windows(*args, **kwargs): check_and_raise() return func(*args, **kwargs) return _wrap_skip_if_on_windows else: check_and_raise()
[docs]def skip_if_root(func=None): """Skip test if uid == 0. Note that on Windows (or anywhere else `os.geteuid` is not available) the test is _not_ skipped. """ check_not_generatorfunction(func) def check_and_raise(): if hasattr(os, "geteuid") and os.geteuid() == 0: raise SkipTest("Skipping: test assumptions fail under root") if func: @wraps(func) @attr('skip_if_root') def _wrap_skip_if_root(*args, **kwargs): check_and_raise() return func(*args, **kwargs) return _wrap_skip_if_root else: check_and_raise()
[docs]@optional_args def skip_if(func, cond=True, msg=None, method='raise'): """Skip test for specific condition Parameters ---------- cond: bool condition on which to skip msg: str message to print if skipping method: str either 'raise' or 'pass'. Whether to skip by raising `SkipTest` or by just proceeding and simply not calling the decorated function. This is particularly meant to be used, when decorating single assertions in a test with method='pass' in order to not skip the entire test, but just that assertion. """ check_not_generatorfunction(func) @wraps(func) def _wrap_skip_if(*args, **kwargs): if cond: if method == 'raise': raise SkipTest(msg if msg else "condition was True") elif method == 'pass': print(msg if msg else "condition was True") return return func(*args, **kwargs) return _wrap_skip_if
[docs]def skip_ssh(func): """Skips SSH tests if on windows or if environment variable DATALAD_TESTS_SSH was not set """ check_not_generatorfunction(func) @wraps(func) @attr('skip_ssh') def _wrap_skip_ssh(*args, **kwargs): test_ssh = dl_cfg.get("datalad.tests.ssh", '') if not test_ssh or test_ssh in ('0', 'false', 'no'): raise SkipTest("Run this test by setting DATALAD_TESTS_SSH") return func(*args, **kwargs) return _wrap_skip_ssh
[docs]def skip_nomultiplex_ssh(func): """Skips SSH tests if default connection/manager does not support multiplexing e.g. currently on windows or if set via datalad.ssh.multiplex-connections config variable """ check_not_generatorfunction(func) from ..support.sshconnector import ( MultiplexSSHManager, SSHManager, ) @wraps(func) @attr('skip_nomultiplex_ssh') @skip_ssh def _wrap_skip_nomultiplex_ssh(*args, **kwargs): if SSHManager is not MultiplexSSHManager: raise SkipTest("SSH without multiplexing is used") return func(*args, **kwargs) return _wrap_skip_nomultiplex_ssh
# # Addition "checkers" # import os from datalad.distribution.dataset import Dataset from datalad.support.annexrepo import ( AnnexRepo, FileNotInAnnexError, ) from datalad.support.gitrepo import GitRepo from ..utils import ( chpwd, getpwd, )
[docs]def ok_clean_git(path, annex=None, index_modified=[], untracked=[]): """Obsolete test helper. Use assert_repo_status() instead. Still maps a few common cases to the new helper, to ease transition in extensions. """ kwargs = {} if index_modified: kwargs['modified'] = index_modified if untracked: kwargs['untracked'] = untracked assert_repo_status( path, annex=annex, **kwargs, )
[docs]def ok_file_under_git(path, filename=None, annexed=False): """Test if file is present and under git/annex control If relative path provided, then test from current directory """ annex, file_repo_path, filename, path, repo = _prep_file_under_git(path, filename) assert_in(file_repo_path, repo.get_indexed_files()) # file is known to Git if annex: in_annex = 'key' in repo.get_file_annexinfo(file_repo_path) else: in_annex = False assert(annexed == in_annex)
[docs]def put_file_under_git(path, filename=None, content=None, annexed=False): """Place file under git/annex and return used Repo """ annex, file_repo_path, filename, path, repo = _prep_file_under_git(path, filename) if content is None: content = "" with open(opj(repo.path, file_repo_path), 'w') as f_: f_.write(content) if annexed: if not isinstance(repo, AnnexRepo): repo = AnnexRepo(repo.path) repo.add(file_repo_path) else: repo.add(file_repo_path, git=True) repo.commit(_datalad_msg=True) ok_file_under_git(repo.path, file_repo_path, annexed) return repo
def _prep_file_under_git(path, filename): """Get instance of the repository for the given filename Helper to be used by few functions """ path = Path(path) if filename is None: # path provides the path and the name filename = Path(path.name) path = path.parent else: filename = Path(filename) ds = Dataset(utils.get_dataset_root(path)) return isinstance(ds.repo, AnnexRepo), \ str(path.absolute().relative_to(ds.path) / filename) \ if not filename.is_absolute() \ else str(filename.relative_to(ds.pathobj)), \ filename, \ str(path), \ ds.repo
[docs]def get_annexstatus(ds, paths=None): """Report a status for annexed contents. Assembles states for git content info, amended with annex info on 'HEAD' (to get the last committed stage and with it possibly vanished content), and lastly annex info wrt to the present worktree, to also get info on added/staged content this fuses the info reported from - git ls-files - git annex findref HEAD - git annex find --include '*'""" info = ds.get_content_annexinfo( paths=paths, eval_availability=False, init=ds.get_content_annexinfo( paths=paths, ref='HEAD', eval_availability=False, init=ds.status( paths=paths, eval_submodule_state='full') ) ) ds._mark_content_availability(info) return info
# # Helpers to test symlinks #
[docs]def ok_startswith(s, prefix): ok_(s.startswith(prefix), msg="String %r doesn't start with %r" % (s, prefix))
[docs]def ok_endswith(s, suffix): ok_(s.endswith(suffix), msg="String %r doesn't end with %r" % (s, suffix))
[docs]def nok_startswith(s, prefix): assert_false(s.startswith(prefix), msg="String %r starts with %r" % (s, prefix))
[docs]def ok_git_config_not_empty(ar): """Helper to verify that nothing rewritten the config file""" # TODO: we don't support bare -- do we? assert_true(os.stat(opj(ar.path, '.git', 'config')).st_size)
[docs]def ok_annex_get(ar, files, network=True): """Helper to run .get decorated checking for correct operation get passes through stderr from the ar to the user, which pollutes screen while running tests Note: Currently not true anymore, since usage of --json disables progressbars """ ok_git_config_not_empty(ar) # we should be working in already inited repo etc with swallow_outputs() as cmo: ar.get(files) # verify that load was fetched ok_git_config_not_empty(ar) # whatever we do shouldn't destroy the config file has_content = ar.file_has_content(files) if isinstance(has_content, bool): ok_(has_content) else: ok_(all(has_content))
[docs]def ok_generator(gen): assert_true(inspect.isgenerator(gen), msg="%s is not a generator" % gen)
assert_is_generator = ok_generator # just an alias
[docs]def ok_archives_caches(repopath, n=1, persistent=None): """Given a path to repository verify number of archives Parameters ---------- repopath : str Path to the repository n : int, optional Number of archives directories to expect persistent: bool or None, optional If None -- both persistent and not count. """ # looking into subdirectories glob_ptn = opj(repopath, ARCHIVES_TEMP_DIR + {None: '*', True: '', False: '-*'}[persistent], '*') dirs = glob.glob(glob_ptn) n2 = n * 2 # per each directory we should have a .stamp file assert_equal(len(dirs), n2, msg="Found following dirs when needed %d of them: %s" % (n2, dirs))
[docs]def ok_exists(path): assert Path(path).exists(), 'path %s does not exist (or dangling symlink)' % path
[docs]def ok_file_has_content(path, content, strip=False, re_=False, decompress=False, **kwargs): """Verify that file exists and has expected content""" path = Path(path) ok_exists(path) if decompress: if path.suffix == '.gz': open_func = gzip.open elif path.suffix in ('.xz', '.lzma'): open_func = lzma.open else: raise NotImplementedError("Don't know how to decompress %s" % path) else: open_func = open with open_func(str(path), 'rb') as f: file_content = f.read() if isinstance(content, str): file_content = ensure_unicode(file_content) if os.linesep != '\n': # for consistent comparisons etc. Apparently when reading in `b` mode # on Windows we would also get \r # https://github.com/datalad/datalad/pull/3049#issuecomment-444128715 file_content = file_content.replace(os.linesep, '\n') if strip: file_content = file_content.strip() if re_: assert_re_in(content, file_content, **kwargs) else: assert_equal(content, file_content, **kwargs)
# # Decorators #
[docs]@optional_args def with_tree(t, tree=None, archives_leading_dir=True, delete=True, **tkwargs): @wraps(t) def _wrap_with_tree(*arg, **kw): if 'dir' not in tkwargs.keys(): # if not specified otherwise, respect datalad.tests.temp.dir config # as this is a test helper tkwargs['dir'] = dl_cfg.get("datalad.tests.temp.dir") tkwargs_ = get_tempfile_kwargs(tkwargs, prefix="tree", wrapped=t) d = tempfile.mkdtemp(**tkwargs_) create_tree(d, tree, archives_leading_dir=archives_leading_dir) try: return t(*(arg + (d,)), **kw) finally: if delete: rmtemp(d) return _wrap_with_tree
lgr = logging.getLogger('datalad.tests')
[docs]class SilentHTTPHandler(SimpleHTTPRequestHandler): """A little adapter to silence the handler """ def __init__(self, *args, **kwargs): self._silent = lgr.getEffectiveLevel() > logging.DEBUG SimpleHTTPRequestHandler.__init__(self, *args, **kwargs)
[docs] def log_message(self, format, *args): if self._silent: return lgr.debug("HTTP: " + format, *args)
def _multiproc_serve_path_via_http( hostname, path_to_serve_from, queue, use_ssl=False, auth=None): # pragma: no cover handler = SilentHTTPHandler if auth: # to-be-expected key for basic auth auth_test = (b'Basic ' + base64.b64encode( bytes('%s:%s' % auth, 'utf-8'))).decode('utf-8') # ad-hoc basic-auth handler class BasicAuthHandler(SilentHTTPHandler): def do_HEAD(self, authenticated): if authenticated: self.send_response(200) else: self.send_response(401) self.send_header( 'WWW-Authenticate', 'Basic realm=\"Protected\"') self.send_header('content-type', 'text/html') self.end_headers() def do_GET(self): if self.headers.get('Authorization') == auth_test: super().do_GET() else: self.do_HEAD(False) self.wfile.write(bytes('Auth failed', 'utf-8')) handler = BasicAuthHandler chpwd(path_to_serve_from) httpd = HTTPServer((hostname, 0), handler) if use_ssl: ca_dir = Path(__file__).parent / 'ca' ssl_key = ca_dir / 'certificate-key.pem' ssl_cert = ca_dir / 'certificate-pub.pem' if any(not p.exists for p in (ssl_key, ssl_cert)): raise RuntimeError( 'SSL requested, but no key/cert file combination can be ' f'located under {ca_dir}') # turn on SSL context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER) context.load_cert_chain(str(ssl_cert), str(ssl_key)) httpd.socket = context.wrap_socket ( httpd.socket, server_side=True) queue.put(httpd.server_port) httpd.serve_forever()
[docs]class HTTPPath(object): """Serve the content of a path via an HTTP URL. This class can be used as a context manager, in which case it returns the URL. Alternatively, the `start` and `stop` methods can be called directly. Parameters ---------- path : str Directory with content to serve. use_ssl : bool auth : tuple Username, password """ def __init__(self, path, use_ssl=False, auth=None): self.path = path self.url = None self._env_patch = None self._mproc = None self.use_ssl = use_ssl self.auth = auth def __enter__(self): self.start() return self.url def __exit__(self, *args): self.stop()
[docs] def start(self): """Start serving `path` via HTTP. """ # There is a problem with Haskell on wheezy trying to # fetch via IPv6 whenever there is a ::1 localhost entry in # /etc/hosts. Apparently fixing that docker image reliably # is not that straightforward, although see # http://jasonincode.com/customizing-hosts-file-in-docker/ # so we just force to use 127.0.0.1 while on wheezy #hostname = '127.0.0.1' if on_debian_wheezy else 'localhost' if self.use_ssl: # we cannot use IPs with SSL certificates hostname = 'localhost' else: hostname = '127.0.0.1' queue = multiprocessing.Queue() self._mproc = multiprocessing.Process( target=_multiproc_serve_path_via_http, args=(hostname, self.path, queue), kwargs=dict(use_ssl=self.use_ssl, auth=self.auth)) self._mproc.start() try: port = queue.get(timeout=300) except multiprocessing.queues.Empty as e: if self.use_ssl: raise SkipTest('No working SSL support') from e else: raise self.url = 'http{}://{}:{}/'.format( 's' if self.use_ssl else '', hostname, port) lgr.debug("HTTP: serving %s under %s", self.path, self.url) # Such tests don't require real network so if http_proxy settings were # provided, we remove them from the env for the duration of this run env = os.environ.copy() if self.use_ssl: env.pop('https_proxy', None) env['REQUESTS_CA_BUNDLE'] = str( Path(__file__).parent / 'ca' / 'ca_bundle.pem') else: env.pop('http_proxy', None) self._env_patch = patch.dict('os.environ', env, clear=True) self._env_patch.start() if self.use_ssl: # verify that the SSL/cert setup is functional, if not skip the # test # python-requests does its own thing re root CA trust # if this fails, check datalad/tests/ca/prov.sh for ca_bundle try: import requests from requests.auth import HTTPBasicAuth r = requests.get( self.url, verify=True, auth=HTTPBasicAuth(*self.auth) if self.auth else None) r.raise_for_status() # be robust and skip if anything goes wrong, rather than just a # particular SSL issue #except requests.exceptions.SSLError as e: except Exception as e: self.stop() raise SkipTest('No working HTTPS setup') from e # now verify that the stdlib tooling also works # if this fails, check datalad/tests/ca/prov.sh # for info on deploying a datalad-root.crt from urllib.request import ( Request, urlopen, ) try: req = Request(self.url) if self.auth: req.add_header( "Authorization", b"Basic " + base64.standard_b64encode( '{0}:{1}'.format(*self.auth).encode('utf-8'))) urlopen(req) # be robust and skip if anything goes wrong, rather than just a # particular SSL issue #except URLError as e: except Exception as e: self.stop() raise SkipTest('No working HTTPS setup') from e
[docs] def stop(self): """Stop serving `path`. """ lgr.debug("HTTP: stopping server under %s", self.path) self._env_patch.stop() self._mproc.terminate()
[docs]@optional_args def serve_path_via_http(tfunc, *targs, use_ssl=False, auth=None): """Decorator which serves content of a directory via http url Parameters ---------- path : str Directory with content to serve. use_ssl : bool Flag whether to set up SSL encryption and return a HTTPS URL. This require a valid certificate setup (which is tested for proper function) or it will cause a SkipTest to be raised. auth : tuple or None If a (username, password) tuple is given, the server access will be protected via HTTP basic auth. """ @wraps(tfunc) @attr('serve_path_via_http') def _wrap_serve_path_via_http(*args, **kwargs): if targs: # if a path is passed into serve_path_via_http, then it's in targs assert len(targs) == 1 path = targs[0] elif len(args) > 1: args, path = args[:-1], args[-1] else: args, path = (), args[0] with HTTPPath(path, use_ssl=use_ssl, auth=auth) as url: return tfunc(*(args + (path, url)), **kwargs) return _wrap_serve_path_via_http
[docs]@optional_args def with_memory_keyring(t): """Decorator to use non-persistant MemoryKeyring instance """ @wraps(t) @attr('with_memory_keyring') def _wrap_with_memory_keyring(*args, **kwargs): keyring = MemoryKeyring() with patch("datalad.downloaders.credentials.keyring_", keyring): return t(*(args + (keyring,)), **kwargs) return _wrap_with_memory_keyring
[docs]@optional_args def without_http_proxy(tfunc): """Decorator to remove http*_proxy env variables for the duration of the test """ @wraps(tfunc) @attr('without_http_proxy') def _wrap_without_http_proxy(*args, **kwargs): if on_windows: raise SkipTest('Unclear why this is not working on windows') # Such tests don't require real network so if http_proxy settings were # provided, we remove them from the env for the duration of this run env = os.environ.copy() env.pop('http_proxy', None) env.pop('https_proxy', None) with patch.dict('os.environ', env, clear=True): return tfunc(*args, **kwargs) return _wrap_without_http_proxy
[docs]@borrowkwargs(methodname=make_tempfile) @optional_args def with_tempfile(t, **tkwargs): """Decorator function to provide a temporary file name and remove it at the end Parameters ---------- To change the used directory without providing keyword argument 'dir' set DATALAD_TESTS_TEMP_DIR. Examples -------- :: @with_tempfile def test_write(tfile): open(tfile, 'w').write('silly test') """ @wraps(t) def _wrap_with_tempfile(*arg, **kw): if 'dir' not in tkwargs.keys(): # if not specified otherwise, respect datalad.tests.temp.dir config # as this is a test helper tkwargs['dir'] = dl_cfg.get("datalad.tests.temp.dir") with make_tempfile(wrapped=t, **tkwargs) as filename: return t(*(arg + (filename,)), **kw) return _wrap_with_tempfile
# ### ### # START known failure decorators # ### ###
[docs]def probe_known_failure(func): """Test decorator allowing the test to pass when it fails and vice versa Setting config datalad.tests.knownfailures.probe to True tests, whether or not the test is still failing. If it's not, an AssertionError is raised in order to indicate that the reason for failure seems to be gone. """ @wraps(func) @attr('probe_known_failure') def _wrap_probe_known_failure(*args, **kwargs): if dl_cfg.obtain("datalad.tests.knownfailures.probe"): assert_raises(Exception, func, *args, **kwargs) # marked as known failure # Note: Since assert_raises lacks a `msg` argument, a comment # in the same line is helpful to determine what's going on whenever # this assertion fails and we see a trace back. Otherwise that line # wouldn't be very telling. else: return func(*args, **kwargs) return _wrap_probe_known_failure
[docs]@optional_args def skip_known_failure(func, method='raise'): """Test decorator allowing to skip a test that is known to fail Setting config datalad.tests.knownfailures.skip to a bool enables/disables skipping. """ @skip_if(cond=dl_cfg.obtain("datalad.tests.knownfailures.skip"), msg="Skip test known to fail", method=method) @wraps(func) @attr('skip_known_failure') def _wrap_skip_known_failure(*args, **kwargs): return func(*args, **kwargs) return _wrap_skip_known_failure
[docs]def known_failure(func): """Test decorator marking a test as known to fail This combines `probe_known_failure` and `skip_known_failure` giving the skipping precedence over the probing. """ @skip_known_failure @probe_known_failure @wraps(func) @attr('known_failure') def _wrap_known_failure(*args, **kwargs): return func(*args, **kwargs) return _wrap_known_failure
[docs]def known_failure_direct_mode(func): """DEPRECATED. Stop using. Does nothing Test decorator marking a test as known to fail in a direct mode test run If datalad.repo.direct is set to True behaves like `known_failure`. Otherwise the original (undecorated) function is returned. """ # TODO: consider adopting nibabel/deprecated.py nibabel/deprecator.py # mechanism to consistently deprecate functionality and ensure they are # displayed. # Since 2.7 Deprecation warnings aren't displayed by default # and thus kinda pointless to issue a warning here, so we will just log msg = "Direct mode support is deprecated, so no point in using " \ "@known_failure_direct_mode for %r since glorious future " \ "DataLad 0.12" % func.__name__ lgr.warning(msg) return func
[docs]def known_failure_windows(func): """Test decorator marking a test as known to fail on windows On Windows behaves like `known_failure`. Otherwise the original (undecorated) function is returned. """ if on_windows: @known_failure @wraps(func) @attr('known_failure_windows') @attr('windows') def dm_func(*args, **kwargs): return func(*args, **kwargs) return dm_func return func
[docs]def known_failure_githubci_win(func): """Test decorator for a known test failure on Github's Windows CI """ if 'GITHUB_WORKFLOW' in os.environ and on_windows: @known_failure @wraps(func) @attr('known_failure_githubci_win') @attr('githubci_win') def dm_func(*args, **kwargs): return func(*args, **kwargs) return dm_func return func
[docs]def known_failure_githubci_osx(func): """Test decorator for a known test failure on Github's macOS CI """ if 'GITHUB_WORKFLOW' in os.environ and on_osx: @known_failure @wraps(func) @attr('known_failure_githubci_osx') @attr('githubci_osx') def dm_func(*args, **kwargs): return func(*args, **kwargs) return dm_func return func
[docs]def known_failure_osx(func): """Test decorator for a known test failure on macOS """ if on_osx: @known_failure @wraps(func) @attr('known_failure_osx') @attr('osx') def dm_func(*args, **kwargs): return func(*args, **kwargs) return dm_func return func
# ### ### # END known failure decorators # ### ### def _get_resolved_flavors(flavors): #flavors_ = (['local', 'clone'] + (['local-url'] if not on_windows else [])) \ # if flavors == 'auto' else flavors flavors_ = (['local', 'clone', 'local-url', 'network'] if not on_windows else ['network', 'network-clone']) \ if flavors == 'auto' else flavors if not isinstance(flavors_, list): flavors_ = [flavors_] if dl_cfg.get('datalad.tests.nonetwork'): flavors_ = [x for x in flavors_ if not x.startswith('network')] return flavors_
[docs]def clone_url(url): runner = GitWitlessRunner() tdir = tempfile.mkdtemp(**get_tempfile_kwargs( {'dir': dl_cfg.get("datalad.tests.temp.dir")}, prefix='clone_url')) runner.run(["git", "clone", url, tdir], protocol=KillOutput) if GitRepo(tdir).is_with_annex(): AnnexRepo(tdir, init=True) _TEMP_PATHS_CLONES.add(tdir) return tdir
local_testrepo_flavors = ['local'] # 'local-url' _TESTREPOS = None def _get_testrepos_uris(regex, flavors): global _TESTREPOS # we should instantiate those whenever test repos actually asked for # TODO: just absorb all this lazy construction within some class if not _TESTREPOS: from .utils_testrepos import ( BasicAnnexTestRepo, BasicGitTestRepo, InnerSubmodule, NestedDataset, SubmoduleDataset, ) _basic_annex_test_repo = BasicAnnexTestRepo() _basic_git_test_repo = BasicGitTestRepo() _submodule_annex_test_repo = SubmoduleDataset() _nested_submodule_annex_test_repo = NestedDataset() _inner_submodule_annex_test_repo = InnerSubmodule() _TESTREPOS = {'basic_annex': {'network': 'https://github.com/datalad/testrepo--basic--r1', 'local': _basic_annex_test_repo.path, 'local-url': _basic_annex_test_repo.url}, 'basic_git': {'local': _basic_git_test_repo.path, 'local-url': _basic_git_test_repo.url}, 'submodule_annex': {'local': _submodule_annex_test_repo.path, 'local-url': _submodule_annex_test_repo.url}, 'nested_submodule_annex': {'local': _nested_submodule_annex_test_repo.path, 'local-url': _nested_submodule_annex_test_repo.url}, # TODO: append 'annex' to the name: # Currently doesn't work with some annex tests, despite # working manually. So, figure out how the tests' setup # messes things up with this one. 'inner_submodule': {'local': _inner_submodule_annex_test_repo.path, 'local-url': _inner_submodule_annex_test_repo.url} } # assure that now we do have those test repos created -- delayed # their creation until actually used _basic_annex_test_repo.create() _basic_git_test_repo.create() _submodule_annex_test_repo.create() _nested_submodule_annex_test_repo.create() _inner_submodule_annex_test_repo.create() uris = [] for name, spec in _TESTREPOS.items(): if not re.match(regex, name): continue uris += [spec[x] for x in set(spec.keys()).intersection(flavors)] # additional flavors which might have not been if 'clone' in flavors and 'clone' not in spec: uris.append(clone_url(spec['local'])) if 'network-clone' in flavors \ and 'network' in spec \ and 'network-clone' not in spec: uris.append(clone_url(spec['network'])) return uris
[docs]@optional_args def with_testrepos(t, regex='.*', flavors='auto', skip=False, count=None): """Decorator to provide a local/remote test repository All tests under datalad/tests/testrepos are stored in two-level hierarchy, where top-level name describes nature/identifier of the test repository, and there could be multiple instances (e.g. generated differently) of the same "content" Parameters ---------- regex : string, optional Regex to select which test repos to use flavors : {'auto', 'local', 'local-url', 'clone', 'network', 'network-clone'} or list of thereof, optional What URIs to provide. E.g. 'local' would just provide path to the repository, while 'network' would provide url of the remote location available on Internet containing the test repository. 'clone' would clone repository first to a temporary location. 'network-clone' would first clone from the network location. 'auto' would include the list of appropriate ones (e.g., no 'network*' flavors if network tests are "forbidden"). count: int, optional If specified, only up to that number of repositories to test with Examples -------- >>> from datalad.tests.utils import with_testrepos >>> @with_testrepos('basic_annex') ... def test_write(repo): ... assert(os.path.exists(os.path.join(repo, '.git', 'annex'))) """ @wraps(t) @attr('with_testrepos') def _wrap_with_testrepos(*arg, **kw): # addurls with our generated file:// URLs doesn't work on appveyor # https://ci.appveyor.com/project/mih/datalad/builds/29841505/job/330rwn2a3cvtrakj #if 'APPVEYOR' in os.environ: # raise SkipTest("Testrepo setup is broken on AppVeyor") # TODO: would need to either avoid this "decorator" approach for # parametric tests or again aggregate failures like sweepargs does flavors_ = _get_resolved_flavors(flavors) testrepos_uris = _get_testrepos_uris(regex, flavors_) # we should always have at least one repo to test on, unless explicitly only # network was requested by we are running without networked tests if not (dl_cfg.get('datalad.tests.nonetwork') and flavors == ['network']): assert(testrepos_uris) else: if not testrepos_uris: raise SkipTest("No non-networked repos to test on") fake_dates = dl_cfg.get("datalad.fake-dates") ntested = 0 for uri in testrepos_uris: if count and ntested >= count: break ntested += 1 if __debug__: lgr.debug('Running %s on %s', t.__name__, uri) try: t(*(arg + (uri,)), **kw) finally: # The is_explicit_path check is needed because it may be a URL, # but check_dates needs a local path or GitRepo object. if fake_dates and is_explicit_path(uri): from ..support.repodates import check_dates assert_false( check_dates(uri, annex="tree")["objects"]) if uri in _TEMP_PATHS_CLONES: _TEMP_PATHS_CLONES.discard(uri) rmtemp(uri) pass # might need to provide additional handling so, handle return _wrap_with_testrepos
with_testrepos.__test__ = False
[docs]@optional_args def with_sameas_remote(func, autoenabled=False): """Provide a repository with a git-annex sameas remote configured. The repository will have two special remotes: r_dir (type=directory) and r_rsync (type=rsync). The rsync remote will be configured with --sameas=r_dir, and autoenabled if `autoenabled` is true. """ from datalad.support.annexrepo import AnnexRepo from datalad.support.exceptions import CommandError @wraps(func) @attr('with_sameas_remotes') @skip_if_on_windows @skip_ssh @with_tempfile(mkdir=True) @with_tempfile(mkdir=True) def _wrap_with_sameas_remote(*args, **kwargs): # With git-annex's 8.20200522-77-g1f2e2d15e, transferring from an rsync # special remote hangs on Xenial. This is likely due to an interaction # with an older rsync or openssh version. Use openssh as a rough # indicator. See # https://git-annex.branchable.com/bugs/Recent_hang_with_rsync_remote_with_older_systems___40__Xenial__44___Jessie__41__/ if external_versions['cmd:system-ssh'] < '7.4' and \ '8.20200522' < external_versions['cmd:annex'] < '8.20200720': raise SkipTest("Test known to hang") sr_path, repo_path = args[-2:] fn_args = args[:-2] repo = AnnexRepo(repo_path) repo.init_remote("r_dir", options=["type=directory", "encryption=none", "directory=" + sr_path]) options = ["type=rsync", "rsyncurl=datalad-test:" + sr_path] if autoenabled: options.append("autoenable=true") options.append("--sameas=r_dir") repo.init_remote("r_rsync", options=options) return func(*(fn_args + (repo,)), **kwargs) return _wrap_with_sameas_remote
[docs]@optional_args def with_fake_cookies_db(func, cookies={}): """mock original cookies db with a fake one for the duration of the test """ from ..support.cookies import cookies_db @wraps(func) @attr('with_fake_cookies_db') def _wrap_with_fake_cookies_db(*args, **kwargs): try: orig_cookies_db = cookies_db._cookies_db cookies_db._cookies_db = cookies.copy() return func(*args, **kwargs) finally: cookies_db._cookies_db = orig_cookies_db return _wrap_with_fake_cookies_db
[docs]@optional_args def assert_cwd_unchanged(func, ok_to_chdir=False): """Decorator to test whether the current working directory remains unchanged Parameters ---------- ok_to_chdir: bool, optional If True, allow to chdir, so this decorator would not then raise exception if chdir'ed but only return to original directory """ @wraps(func) def _wrap_assert_cwd_unchanged(*args, **kwargs): cwd_before = os.getcwd() pwd_before = getpwd() exc_info = None # record previous state of PWD handling utils_pwd_mode = utils._pwd_mode try: ret = func(*args, **kwargs) except: exc_info = sys.exc_info() finally: utils._pwd_mode = utils_pwd_mode try: cwd_after = os.getcwd() except OSError as e: lgr.warning("Failed to getcwd: %s" % e) cwd_after = None if cwd_after != cwd_before: chpwd(pwd_before) # Above chpwd could also trigger the change of _pwd_mode, so we # would need to reset it again since we know that it is all kosher utils._pwd_mode = utils_pwd_mode if not ok_to_chdir: lgr.warning( "%s changed cwd to %s. Mitigating and changing back to %s" % (func, cwd_after, pwd_before)) # If there was already exception raised, we better reraise # that one since it must be more important, so not masking it # here with our assertion if exc_info is None: assert_equal(cwd_before, cwd_after, "CWD changed from %s to %s" % (cwd_before, cwd_after)) if exc_info is not None: raise exc_info[1] return ret return _wrap_assert_cwd_unchanged
[docs]@optional_args def run_under_dir(func, newdir='.'): """Decorator to run tests under another directory It is somewhat ugly since we can't really chdir back to a directory which had a symlink in its path. So using this decorator has potential to move entire testing run under the dereferenced directory name -- sideeffect. The only way would be to instruct testing framework (i.e. nose in our case ATM) to run a test by creating a new process with a new cwd """ @wraps(func) def _wrap_run_under_dir(*args, **kwargs): pwd_before = getpwd() try: chpwd(newdir) func(*args, **kwargs) finally: chpwd(pwd_before) return _wrap_run_under_dir
[docs]def assert_re_in(regex, c, flags=0, match=True, msg=None): """Assert that container (list, str, etc) contains entry matching the regex """ if not isinstance(c, (list, tuple)): c = [c] for e in c: if (re.match if match else re.search)(regex, e, flags=flags): return raise AssertionError( msg or "Not a single entry matched %r in %r" % (regex, c) )
[docs]def assert_dict_equal(d1, d2): msgs = [] if set(d1).difference(d2): msgs.append(" keys in the first dict but not in the second: %s" % list(set(d1).difference(d2))) if set(d2).difference(d1): msgs.append(" keys in the second dict but not in the first: %s" % list(set(d2).difference(d1))) for k in set(d1).intersection(d2): same = True try: if isinstance(d1[k], str): # do not compare types for string types to avoid all the hassle # with the distinction of str and unicode in PY3, and simple # test for equality same = bool(d1[k] == d2[k]) else: same = type(d1[k]) == type(d2[k]) and bool(d1[k] == d2[k]) except: # if comparison or conversion to bool (e.g. with numpy arrays) fails same = False if not same: msgs.append(" [%r] differs: %r != %r" % (k, d1[k], d2[k])) if len(msgs) > 10: msgs.append("and more") break if msgs: raise AssertionError("dicts differ:\n%s" % "\n".join(msgs)) # do generic comparison just in case we screwed up to detect difference correctly above eq_(d1, d2)
[docs]def assert_str_equal(s1, s2): """Helper to compare two lines""" diff = list(unified_diff(s1.splitlines(), s2.splitlines())) assert not diff, '\n'.join(diff) assert_equal(s1, s2)
[docs]def assert_status(label, results): """Verify that each status dict in the results has a given status label `label` can be a sequence, in which case status must be one of the items in this sequence. """ label = ensure_list(label) results = ensure_list(results) for i, r in enumerate(results): try: assert_in('status', r) assert_in(r['status'], label) except AssertionError: raise AssertionError('Test {}/{}: expected status {} not found in:\n{}'.format( i + 1, len(results), label, dumps(results, indent=1, default=lambda x: str(x))))
[docs]def assert_message(message, results): """Verify that each status dict in the results has a message This only tests the message template string, and not a formatted message with args expanded. """ for r in ensure_list(results): assert_in('message', r) m = r['message'][0] if isinstance(r['message'], tuple) else r['message'] assert_equal(m, message)
def _format_res(x): return textwrap.indent( dumps(x, indent=1, default=str, sort_keys=True), prefix=" ")
[docs]def assert_result_count(results, n, **kwargs): """Verify specific number of results (matching criteria, if any)""" count = 0 results = ensure_list(results) for r in results: if not len(kwargs): count += 1 elif all(k in r and r[k] == v for k, v in kwargs.items()): count += 1 if not n == count: raise AssertionError( 'Got {} instead of {} expected results matching\n{}\nInspected {} record(s):\n{}'.format( count, n, _format_res(kwargs), len(results), _format_res(results)))
def _check_results_in(should_contain, results, **kwargs): found = False for r in ensure_list(results): if all(k in r and r[k] == v for k, v in kwargs.items()): found = True break if found ^ should_contain: if should_contain: msg = "Desired result\n{}\nnot found among\n{}" else: msg = "Result\n{}\nunexpectedly found among\n{}" raise AssertionError(msg.format(_format_res(kwargs), _format_res(results)))
[docs]def assert_in_results(results, **kwargs): """Verify that the particular combination of keys and values is found in one of the results""" _check_results_in(True, results, **kwargs)
[docs]def assert_not_in_results(results, **kwargs): """Verify that the particular combination of keys and values is not in any of the results""" _check_results_in(False, results, **kwargs)
[docs]def assert_result_values_equal(results, prop, values): """Verify that the values of all results for a given key in the status dicts match the given sequence""" assert_equal( [r[prop] for r in results], values)
[docs]def assert_result_values_cond(results, prop, cond): """Verify that the values of all results for a given key in the status dicts fulfill condition `cond`. Parameters ---------- results: prop: str cond: callable """ for r in ensure_list(results): ok_(cond(r[prop]), msg="r[{prop}]: {value}".format(prop=prop, value=r[prop]))
[docs]def ignore_nose_capturing_stdout(func): """DEPRECATED and will be removed soon. Does nothing! Originally was intended as a decorator workaround for nose's behaviour with redirecting sys.stdout, but now we monkey patch nose now so no test should no longer be skipped. See issue reported here: https://code.google.com/p/python-nose/issues/detail?id=243&can=1&sort=-id&colspec=ID%20Type%20Status%20Priority%20Stars%20Milestone%20Owner%20Summary """ lgr.warning( "@ignore_nose_capturing_stdout no longer does anything - nose should " "just be monkey patched in setup_package. {} still has it" .format(func.__name__) ) return func
[docs]@optional_args def with_parametric_batch(t): """Helper to run parametric test with possible combinations of batch and direct """ @wraps(t) def _wrap_with_parametric_batch(): for batch in (False, True): yield t, batch return _wrap_with_parametric_batch
# List of most obscure filenames which might or not be supported by different # filesystems across different OSs. Start with the most obscure OBSCURE_PREFIX = os.getenv('DATALAD_TESTS_OBSCURE_PREFIX', '') # Those will be tried to be added to the base name if filesystem allows OBSCURE_FILENAME_PARTS = [' ', '/', '|', ';', '&', '%b5', '{}', "'", '"'] UNICODE_FILENAME = u"ΔЙקم๗あ" # OSX is exciting -- some I guess FS might be encoding differently from decoding # so Й might get recoded # (ref: https://github.com/datalad/datalad/pull/1921#issuecomment-385809366) if sys.getfilesystemencoding().lower() == 'utf-8': if on_osx: # TODO: figure it really out UNICODE_FILENAME = UNICODE_FILENAME.replace(u"Й", u"") if on_windows: # TODO: really figure out unicode handling on windows UNICODE_FILENAME = '' if UNICODE_FILENAME: OBSCURE_FILENAME_PARTS.append(UNICODE_FILENAME) # space before extension, simple extension and trailing space to finish it up OBSCURE_FILENAME_PARTS += [' ', '.datc', ' ']
[docs]@with_tempfile(mkdir=True) def get_most_obscure_supported_name(tdir, return_candidates=False): """Return the most obscure filename that the filesystem would support under TEMPDIR Parameters ---------- return_candidates: bool, optional if True, return a tuple of (good, candidates) where candidates are "partially" sorted from trickiest considered TODO: we might want to use it as a function where we would provide tdir """ # we need separate good_base so we do not breed leading/trailing spaces initial = good = OBSCURE_PREFIX system = platform.system() OBSCURE_FILENAMES = [] def good_filename(filename): OBSCURE_FILENAMES.append(candidate) try: # Windows seems to not tollerate trailing spaces and # ATM we do not distinguish obscure filename and dirname. # So here we will test for both - being able to create dir # with obscure name and obscure filename under os.mkdir(opj(tdir, filename)) with open(opj(tdir, filename, filename), 'w') as f: f.write("TEST LOAD") return True except: lgr.debug("Filename %r is not supported on %s under %s", filename, system, tdir) return False # incrementally build up the most obscure filename from parts for part in OBSCURE_FILENAME_PARTS: candidate = good + part if good_filename(candidate): good = candidate if good == initial: raise RuntimeError("Could not create any of the files under %s among %s" % (tdir, OBSCURE_FILENAMES)) lgr.debug("Tested %d obscure filename candidates. The winner: %r", len(OBSCURE_FILENAMES), good) if return_candidates: return good, OBSCURE_FILENAMES[::-1] else: return good
OBSCURE_FILENAME, OBSCURE_FILENAMES = get_most_obscure_supported_name(return_candidates=True)
[docs]@optional_args def with_testsui(t, responses=None, interactive=True): """Switch main UI to be 'tests' UI and possibly provide answers to be used""" @wraps(t) def _wrap_with_testsui(*args, **kwargs): from datalad.ui import ui old_backend = ui.backend try: ui.set_backend('tests' if interactive else 'tests-noninteractive') if responses: ui.add_responses(responses) ret = t(*args, **kwargs) if responses: responses_left = ui.get_responses() assert not len(responses_left), "Some responses were left not used: %s" % str(responses_left) return ret finally: ui.set_backend(old_backend) if not interactive and responses is not None: raise ValueError("Non-interactive UI cannot provide responses") return _wrap_with_testsui
with_testsui.__test__ = False
[docs]def assert_no_errors_logged(func, skip_re=None): """Decorator around function to assert that no errors logged during its execution""" @wraps(func) def _wrap_assert_no_errors_logged(*args, **kwargs): with swallow_logs(new_level=logging.ERROR) as cml: out = func(*args, **kwargs) if cml.out: if not (skip_re and re.search(skip_re, cml.out)): raise AssertionError( "Expected no errors to be logged, but log output is %s" % cml.out ) return out return _wrap_assert_no_errors_logged
[docs]def get_mtimes_and_digests(target_path): """Return digests (md5) and mtimes for all the files under target_path""" from datalad.support.digests import Digester from datalad.utils import find_files digester = Digester(['md5']) # bother only with existing ones for this test, i.e. skip annexed files without content target_files = [ f for f in find_files('.*', topdir=target_path, exclude_vcs=False, exclude_datalad=False) if exists(f) ] # let's leave only relative paths for easier analysis target_files_ = [relpath(f, target_path) for f in target_files] digests = {frel: digester(f) for f, frel in zip(target_files, target_files_)} mtimes = {frel: os.stat(f).st_mtime for f, frel in zip(target_files, target_files_)} return digests, mtimes
[docs]def get_datasets_topdir(): """Delayed parsing so it could be monkey patched etc""" from datalad.consts import DATASETS_TOPURL return RI(DATASETS_TOPURL).hostname
[docs]def assert_repo_status(path, annex=None, untracked_mode='normal', **kwargs): """Compare a repo status against (optional) exceptions. Anything file/directory that is not explicitly indicated must have state 'clean', i.e. no modifications and recorded in Git. Parameters ---------- path: str or Repo in case of a str: path to the repository's base dir; Note, that passing a Repo instance prevents detecting annex. This might be useful in case of a non-initialized annex, a GitRepo is pointing to. annex: bool or None explicitly set to True or False to indicate, that an annex is (not) expected; set to None to autodetect, whether there is an annex. Default: None. untracked_mode: {'no', 'normal', 'all'} If and how untracked content is reported. The specification of untracked files that are OK to be found must match this mode. See `Repo.status()` **kwargs Files/directories that are OK to not be in 'clean' state. Each argument must be one of 'added', 'untracked', 'deleted', 'modified' and each value must be a list of filenames (relative to the root of the repository, in POSIX convention). """ r = None if isinstance(path, AnnexRepo): if annex is None: annex = True # if `annex` was set to False, but we find an annex => fail assert_is(annex, True) r = path elif isinstance(path, GitRepo): if annex is None: annex = False # explicitly given GitRepo instance doesn't make sense with # 'annex' True assert_is(annex, False) r = path else: # 'path' is an actual path try: r = AnnexRepo(path, init=False, create=False) if annex is None: annex = True # if `annex` was set to False, but we find an annex => fail assert_is(annex, True) except Exception: # Instantiation failed => no annex try: r = GitRepo(path, init=False, create=False) except Exception: raise AssertionError("Couldn't find an annex or a git " "repository at {}.".format(path)) if annex is None: annex = False # explicitly given GitRepo instance doesn't make sense with # 'annex' True assert_is(annex, False) status = r.status(untracked=untracked_mode) # for any file state that indicates some kind of change (all but 'clean) for state in ('added', 'untracked', 'deleted', 'modified'): oktobefound = sorted(r.pathobj.joinpath(ut.PurePosixPath(p)) for p in kwargs.get(state, [])) state_files = sorted(k for k, v in status.items() if v.get('state', None) == state) eq_(state_files, oktobefound, 'unexpected content of state "%s": %r != %r' % (state, state_files, oktobefound))
[docs]def get_convoluted_situation(path, repocls=AnnexRepo): from datalad.api import create #if 'APPVEYOR' in os.environ: # # issue only happens on appveyor, Python itself implodes # # cannot be reproduced on a real windows box # raise SkipTest( # 'get_convoluted_situation() causes appveyor to crash, ' # 'reason unknown') repo = repocls(path, create=True) # use create(force) to get an ID and config into the empty repo ds = Dataset(path).create(force=True) # base content create_tree( ds.path, { '.gitignore': '*.ignored', 'subdir': { 'file_clean': 'file_clean', 'file_deleted': 'file_deleted', 'file_modified': 'file_clean', }, 'subdir-only-ignored': { '1.ignored': '', }, 'file_clean': 'file_clean', 'file_deleted': 'file_deleted', 'file_staged_deleted': 'file_staged_deleted', 'file_modified': 'file_clean', } ) if isinstance(ds.repo, AnnexRepo): create_tree( ds.path, { 'subdir': { 'file_dropped_clean': 'file_dropped_clean', }, 'file_dropped_clean': 'file_dropped_clean', } ) ds.save() if isinstance(ds.repo, AnnexRepo): # some files straight in git create_tree( ds.path, { 'subdir': { 'file_ingit_clean': 'file_ingit_clean', 'file_ingit_modified': 'file_ingit_clean', }, 'file_ingit_clean': 'file_ingit_clean', 'file_ingit_modified': 'file_ingit_clean', } ) ds.save(to_git=True) ds.drop([ 'file_dropped_clean', opj('subdir', 'file_dropped_clean')], check=False) # clean and proper subdatasets ds.create('subds_clean') ds.create(opj('subdir', 'subds_clean')) ds.create('subds_unavailable_clean') ds.create(opj('subdir', 'subds_unavailable_clean')) # uninstall some subdatasets (still clean) ds.drop([ 'subds_unavailable_clean', opj('subdir', 'subds_unavailable_clean')], what='all', reckless='kill', recursive=True) assert_repo_status(ds.path) # make a dirty subdataset ds.create('subds_modified') ds.create(opj('subds_modified', 'someds')) ds.create(opj('subds_modified', 'someds', 'dirtyds')) # make a subdataset with additional commits ds.create(opj('subdir', 'subds_modified')) pdspath = opj(ds.path, 'subdir', 'subds_modified', 'progressedds') ds.create(pdspath) create_tree( pdspath, {'file_clean': 'file_ingit_clean'} ) Dataset(pdspath).save() assert_repo_status(pdspath) # staged subds, and files create(opj(ds.path, 'subds_added')) # use internal helper to get subdataset into an 'added' state # that would not happen in standard datalad workflows list(ds.repo._save_add_submodules([ds.pathobj / 'subds_added'])) create(opj(ds.path, 'subdir', 'subds_added')) list(ds.repo._save_add_submodules([ds.pathobj / 'subdir' / 'subds_added'])) # some more untracked files create_tree( ds.path, { 'subdir': { 'file_untracked': 'file_untracked', 'file_added': 'file_added', }, 'file_untracked': 'file_untracked', 'file_added': 'file_added', 'dir_untracked': { 'file_untracked': 'file_untracked', }, 'subds_modified': { 'someds': { "dirtyds": { 'file_untracked': 'file_untracked', }, }, }, } ) ds.repo.add(['file_added', opj('subdir', 'file_added')]) # untracked subdatasets create(opj(ds.path, 'subds_untracked')) create(opj(ds.path, 'subdir', 'subds_untracked')) # deleted files os.remove(opj(ds.path, 'file_deleted')) os.remove(opj(ds.path, 'subdir', 'file_deleted')) # staged deletion ds.repo.remove('file_staged_deleted') # modified files if isinstance(ds.repo, AnnexRepo): ds.repo.unlock(['file_modified', opj('subdir', 'file_modified')]) create_tree( ds.path, { 'subdir': { 'file_ingit_modified': 'file_ingit_modified', }, 'file_ingit_modified': 'file_ingit_modified', } ) create_tree( ds.path, { 'subdir': { 'file_modified': 'file_modified', }, 'file_modified': 'file_modified', } ) return ds
[docs]def get_deeply_nested_structure(path): """ Here is what this does (assuming UNIX, locked): | . | ├── directory_untracked | │ └── link2dir -> ../subdir | ├── OBSCURE_FILENAME_file_modified | ├── link2dir -> subdir | ├── link2subdsdir -> subds_modified/subdir | ├── link2subdsroot -> subds_modified | ├── subdir | │ ├── annexed_file.txt -> ../.git/annex/objects/... | │ ├── file_modified | │ ├── git_file.txt | │ └── link2annex_files.txt -> annexed_file.txt | └── subds_modified | ├── link2superdsdir -> ../subdir | ├── subdir | │ └── annexed_file.txt -> ../.git/annex/objects/... | └── subds_lvl1_modified | └── OBSCURE_FILENAME_directory_untracked | └── untracked_file When a system has no symlink support, the link2... components are not included. """ ds = Dataset(path).create() (ds.pathobj / 'subdir').mkdir() (ds.pathobj / 'subdir' / 'annexed_file.txt').write_text(u'dummy') ds.save() (ds.pathobj / 'subdir' / 'git_file.txt').write_text(u'dummy') ds.save(to_git=True) # a subtree of datasets subds = ds.create('subds_modified') # another dataset, plus an additional dir in it ds.create(opj('subds_modified', 'subds_lvl1_modified')) create_tree( ds.path, { 'subdir': { 'file_modified': 'file_modified', }, OBSCURE_FILENAME + u'file_modified_': 'file_modified', } ) create_tree( str(ds.pathobj / 'subds_modified' / 'subds_lvl1_modified'), {OBSCURE_FILENAME + u'_directory_untracked': {"untracked_file": ""}} ) (ut.Path(subds.path) / 'subdir').mkdir() (ut.Path(subds.path) / 'subdir' / 'annexed_file.txt').write_text(u'dummy') subds.save() (ds.pathobj / 'directory_untracked').mkdir() if not has_symlink_capability(): return ds # symlink farm #1 # symlink to annexed file (ds.pathobj / 'subdir' / 'link2annex_files.txt').symlink_to( 'annexed_file.txt') # symlink to directory within the dataset (ds.pathobj / 'link2dir').symlink_to('subdir') # upwards pointing symlink to directory within the same dataset (ds.pathobj / 'directory_untracked' / 'link2dir').symlink_to( opj('..', 'subdir')) # symlink pointing to a subdataset mount in the same dataset (ds.pathobj / 'link2subdsroot').symlink_to('subds_modified') # symlink to a dir in a subdataset (across dataset boundaries) (ds.pathobj / 'link2subdsdir').symlink_to( opj('subds_modified', 'subdir')) # symlink to a dir in a superdataset (across dataset boundaries) (ut.Path(subds.path) / 'link2superdsdir').symlink_to( opj('..', 'subdir')) return ds
[docs]def maybe_adjust_repo(repo): """Put repo into an adjusted branch if it is not already. """ if not repo.is_managed_branch(): repo.call_annex(["upgrade"]) repo.config.reload(force=True) repo.adjust()
_TESTS_ADJUSTED_TMPDIR = None
[docs]def skip_if_adjusted_branch(func): """Skip test if adjusted branch is used by default on TMPDIR file system. """ @wraps(func) @attr('skip_if_adjusted_branch') def _wrap_skip_if_adjusted_branch(*args, **kwargs): global _TESTS_ADJUSTED_TMPDIR if _TESTS_ADJUSTED_TMPDIR is None: @with_tempfile def _check(path): ds = Dataset(path).create(force=True) return ds.repo.is_managed_branch() _TESTS_ADJUSTED_TMPDIR = _check() if _TESTS_ADJUSTED_TMPDIR: raise SkipTest("Test incompatible with adjusted branch default") return func(*args, **kwargs) return _wrap_skip_if_adjusted_branch
[docs]def get_ssh_port(host): """Get port of `host` in ssh_config. Our tests depend on the host being defined in ssh_config, including its port. This method can be used by tests that want to check handling of an explicitly specified Note that if `host` does not match a host in ssh_config, the default value of 22 is returned. Parameters ---------- host : str Returns ------- port (int) Raises ------ SkipTest if port cannot be found. """ out = '' runner = WitlessRunner() try: res = runner.run(["ssh", "-G", host], protocol=StdOutErrCapture) out = res["stdout"] err = res["stderr"] except Exception as exc: err = str(exc) port = None for line in out.splitlines(): if line.startswith("port "): try: port = int(line.split()[1]) except Exception as exc: err = str(exc) break if port is None: raise SkipTest("port for {} could not be determined: {}" .format(host, err)) return port
# # Context Managers #
[docs]def patch_config(vars): """Patch our config with custom settings. Returns mock.patch cm Only the merged configuration from all sources (global, local, dataset) will be patched. Source-constrained patches (e.g. only committed dataset configuration) are not supported. """ return patch.dict(dl_cfg._merged_store, vars)
[docs]@contextmanager def set_date(timestamp): """Temporarily override environment variables for git/git-annex dates. Parameters ---------- timestamp : int Unix timestamp. """ git_ts = "@{} +0000".format(timestamp) with patch.dict("os.environ", {"GIT_COMMITTER_DATE": git_ts, "GIT_AUTHOR_DATE": git_ts, "GIT_ANNEX_VECTOR_CLOCK": str(timestamp), "DATALAD_FAKE__DATES": "0"}): yield
[docs]@contextmanager def set_annex_version(version): """Override the git-annex version. This temporarily masks the git-annex version present in external_versions and make AnnexRepo forget its cached version information. """ from datalad.support.annexrepo import AnnexRepo ar_vers = AnnexRepo.git_annex_version with patch.dict( "datalad.support.annexrepo.external_versions._versions", {"cmd:annex": version}): try: AnnexRepo.git_annex_version = None yield finally: AnnexRepo.git_annex_version = ar_vers
# # Test tags # # To be explicit, and not "loose" some tests due to typos, decided to make # explicit decorators for common types
[docs]def integration(f): """Mark test as an "integration" test which generally is not needed to be run Generally tend to be slower. Should be used in combination with @slow and @turtle if that is the case. """ return attr('integration')(f)
[docs]def slow(f): """Mark test as a slow, although not necessarily integration or usecase test Rule of thumb cut-off to mark as slow is 10 sec """ return attr('slow')(f)
[docs]def turtle(f): """Mark test as very slow, meaning to not run it on Travis due to its time limit Rule of thumb cut-off to mark as turtle is 2 minutes """ return attr('turtle')(f)
[docs]def usecase(f): """Mark test as a usecase user ran into and which (typically) caused bug report to be filed/troubleshooted Should be used in combination with @slow and @turtle if slow. """ return attr('usecase')(f)