Exemple #1
0
def test_add_archive_content(path_orig, url, repo_path):
    with chpwd(repo_path):
        # TODO we need to be able to pass path into add_archive_content
        # We could mock but I mean for the API
        assert_raises(RuntimeError, add_archive_content,
                      "nonexisting.tar.gz")  # no repo yet

        repo = AnnexRepo(repo_path, create=True)
        assert_raises(ValueError, add_archive_content, "nonexisting.tar.gz")
        # we can't add a file from outside the repo ATM
        assert_raises(FileNotInRepositoryError, add_archive_content,
                      opj(path_orig, '1.tar.gz'))

        # Let's add first archive to the repo so we could test
        with swallow_outputs():
            repo.add_urls([opj(url, '1.tar.gz')],
                          options=["--pathdepth", "-1"])
            for s in range(1, 5):
                repo.add_urls([opj(url, '%du/1.tar.gz' % s)],
                              options=["--pathdepth", "-2"])
        repo.commit("added 1.tar.gz")

        key_1tar = repo.get_file_key(
            '1.tar.gz')  # will be used in the test later

        def d1_basic_checks():
            ok_(exists('1'))
            ok_file_under_git('1', '1 f.txt', annexed=True)
            ok_file_under_git(opj('1', 'd', '1d'), annexed=True)
            ok_archives_caches(repo_path, 0)

        # and by default it just does it, everything goes to annex
        repo_ = add_archive_content('1.tar.gz')
        eq_(repo.path, repo_.path)
        d1_basic_checks()

        # If ran again, should proceed just fine since the content is the same so no changes would be made really
        add_archive_content('1.tar.gz')

        # But that other one carries updated file, so should fail due to overwrite
        with assert_raises(RuntimeError) as cme:
            add_archive_content(opj('1u', '1.tar.gz'), use_current_dir=True)

        # TODO: somewhat not precise since we have two possible "already exists"
        # -- in caching and overwrite check
        assert_in("already exists", str(cme.exception))
        # but should do fine if overrides are allowed
        add_archive_content(opj('1u', '1.tar.gz'),
                            existing='overwrite',
                            use_current_dir=True)
        add_archive_content(opj('2u', '1.tar.gz'),
                            existing='archive-suffix',
                            use_current_dir=True)
        add_archive_content(opj('3u', '1.tar.gz'),
                            existing='archive-suffix',
                            use_current_dir=True)
        add_archive_content(opj('4u', '1.tar.gz'),
                            existing='archive-suffix',
                            use_current_dir=True)

        # rudimentary test
        assert_equal(sorted(map(basename, glob(opj(repo_path, '1', '1*')))),
                     ['1 f-1.1.txt', '1 f-1.2.txt', '1 f-1.txt', '1 f.txt'])
        whereis = repo.whereis(glob(opj(repo_path, '1', '1*')))
        # they all must be the same
        assert (all([x == whereis[0] for x in whereis[1:]]))

    # and we should be able to reference it while under subdirectory
    subdir = opj(repo_path, 'subdir')
    with chpwd(subdir, mkdir=True):
        add_archive_content(opj(pardir, '1.tar.gz'), use_current_dir=True)
        d1_basic_checks()
        # or we could keep relative path and also demand to keep the archive prefix
        # while extracting under original (annex root) dir
        add_archive_content(opj(pardir, '1.tar.gz'),
                            add_archive_leading_dir=True)

    with chpwd(opj(repo_path, '1')):
        d1_basic_checks()

    with chpwd(repo_path):
        # test with excludes and renames and annex options
        add_archive_content('1.tar.gz',
                            exclude=['d'],
                            rename=['/ /_', '/^1/2'],
                            annex_options="-c annex.largefiles=exclude=*.txt",
                            delete=True)
        # no conflicts since new name
        ok_file_under_git('2', '1_f.txt', annexed=False)
        assert_false(exists(opj('2', 'd')))
        assert_false(exists('1.tar.gz'))  # delete was in effect

    # now test ability to extract within subdir
    with chpwd(opj(repo_path, 'd1'), mkdir=True):
        # Let's add first archive to the repo so we could test
        # named the same way but different content
        with swallow_outputs():
            repo.add_urls([opj(url, 'd1', '1.tar.gz')],
                          options=["--pathdepth", "-1"],
                          cwd=getpwd())  # invoke under current subdir
        repo.commit("added 1.tar.gz in d1")

        def d2_basic_checks():
            ok_(exists('1'))
            ok_file_under_git('1', '2 f.txt', annexed=True)
            ok_file_under_git(opj('1', 'd2', '2d'), annexed=True)
            ok_archives_caches(repo.path, 0)

        add_archive_content('1.tar.gz')
        d2_basic_checks()

    # in manual tests ran into the situation of inability to obtain on a single run
    # a file from an archive which was coming from a dropped key.  I thought it was
    # tested in custom remote tests, but I guess not sufficiently well enough
    repo.drop(opj('1', '1 f.txt'))  # should be all kosher
    repo.get(opj('1', '1 f.txt'))
    ok_archives_caches(repo.path, 1, persistent=True)
    ok_archives_caches(repo.path, 0, persistent=False)

    repo.drop(opj('1', '1 f.txt'))  # should be all kosher
    repo.drop(key_1tar,
              key=True)  # is available from the URL -- should be kosher
    repo.get(opj('1', '1 f.txt'))  # that what managed to not work

    # TODO: check if persistent archive is there for the 1.tar.gz

    # We should be able to drop everything since available online
    with swallow_outputs():
        clean(dataset=repo.path)
    repo.drop(key_1tar,
              key=True)  # is available from the URL -- should be kosher

    repo.drop(opj('1', '1 f.txt'))  # should be all kosher
    repo.get(opj('1', '1 f.txt'))  # and should be able to get it again

    # bug was that dropping didn't work since archive was dropped first
    repo.call_annex(["drop", "--all"])

    # verify that we can't drop a file if archive key was dropped and online archive was removed or changed size! ;)
    repo.get(key_1tar, key=True)
    unlink(opj(path_orig, '1.tar.gz'))
    with assert_raises(CommandError) as e:
        repo.drop(key_1tar, key=True)
        assert_equal(e.kwargs['stdout_json'][0]['success'], False)
        assert_result_values_cond(
            e.kwargs['stdout_json'], 'note', lambda x:
            '(Use --force to override this check, or adjust numcopies.)' in x)
    assert exists(opj(repo.path, repo.get_contentlocation(key_1tar)))
def test_target_ssh_simple(origin, src_path, target_rootpath):
    port = get_ssh_port("datalad-test")
    # prepare src
    source = install(src_path,
                     source=origin,
                     result_xfm='datasets',
                     return_type='item-or-list')

    target_path = opj(target_rootpath, "basic")
    with swallow_logs(new_level=logging.ERROR) as cml:
        create_sibling(dataset=source,
                       name="local_target",
                       sshurl="ssh://*****:*****@with_testsui(responses=["yes"])
        def interactive_assert_create_sshwebserver():
            assert_create_sshwebserver(
                dataset=source,
                name="local_target",
                sshurl="ssh://datalad-test" + target_path,
                publish_by_default=DEFAULT_BRANCH,
                existing='replace',
                ui=have_webui(),
            )

        interactive_assert_create_sshwebserver()

        eq_("ssh://datalad-test" + urlquote(target_path),
            source.repo.get_remote_url("local_target"))
        ok_(source.repo.get_remote_url("local_target", push=True) is None)

        # ensure target tree actually replaced by source
        assert_false(exists(opj(target_path, 'random')))

        if src_is_annex:
            lclcfg = AnnexRepo(src_path).config
            eq_(lclcfg.get('remote.local_target.annex-ignore'), 'false')
            # valid uuid
            eq_(lclcfg.get('remote.local_target.annex-uuid').count('-'), 4)
            # should be added too, even if URL matches prior state
            eq_(lclcfg.get('remote.local_target.push'), DEFAULT_BRANCH)

        # again, by explicitly passing urls. Since we are on datalad-test, the
        # local path should work:
        cpkwargs = dict(
            dataset=source,
            name="local_target",
            sshurl="ssh://datalad-test",
            target_dir=target_path,
            target_url=target_path,
            target_pushurl="ssh://datalad-test" + target_path,
            ui=have_webui(),
        )

        @with_testsui(responses=['yes'])
        def interactive_assert_create_sshwebserver():
            assert_create_sshwebserver(existing='replace', **cpkwargs)

        interactive_assert_create_sshwebserver()

        if src_is_annex:
            target_description = AnnexRepo(target_path,
                                           create=False).get_description()
            eq_(target_description, target_path)

        eq_(target_path, source.repo.get_remote_url("local_target"))
        eq_("ssh://datalad-test" + target_path,
            source.repo.get_remote_url("local_target", push=True))

        if have_webui():
            from datalad_deprecated.tests.test_create_sibling_webui \
                import assert_publish_with_ui
            assert_publish_with_ui(target_path)

        # now, push should work:
        publish(dataset=source, to="local_target")

        # and we should be able to 'reconfigure'
        def process_digests_mtimes(digests, mtimes):
            # it should have triggered a hook, which would have created log and metadata files
            check_metadata = False
            for part in 'logs', 'metadata':
                metafiles = [
                    k for k in digests
                    if k.startswith(_path_('.git/datalad/%s/' % part))
                ]
                # This is in effect ONLY if we have "compatible" datalad installed on remote
                # end. ATM we don't have easy way to guarantee that AFAIK (yoh),
                # so let's not check/enforce (TODO)
                # assert(len(metafiles) >= 1)  # we might have 2 logs if timestamps do not collide ;)
                # Let's actually do it to some degree
                if part == 'logs':
                    # always should have those:
                    assert (len(metafiles) >= 1)
                    with open(opj(target_path, metafiles[0])) as f:
                        if 'no datalad found' not in f.read():
                            check_metadata = True
                if part == 'metadata':
                    eq_(len(metafiles), bool(check_metadata))
                for f in metafiles:
                    digests.pop(f)
                    mtimes.pop(f)
            # and just pop some leftovers from annex
            # and ignore .git/logs content (gh-5298)
            for f in list(digests):
                if f.startswith('.git/annex/mergedrefs') \
                        or f.startswith('.git/logs/'):
                    digests.pop(f)
                    mtimes.pop(f)

        if not have_webui():
            # the rest of the test assumed that we have uploaded a UI
            return
        orig_digests, orig_mtimes = get_mtimes_and_digests(target_path)
        process_digests_mtimes(orig_digests, orig_mtimes)

        import time
        time.sleep(0.1)  # just so that mtimes change
        assert_create_sshwebserver(existing='reconfigure', **cpkwargs)
        digests, mtimes = get_mtimes_and_digests(target_path)
        process_digests_mtimes(digests, mtimes)

        assert_dict_equal(orig_digests,
                          digests)  # nothing should change in terms of content

        # but some files should have been modified
        modified_files = {
            k
            for k in mtimes if orig_mtimes.get(k, 0) != mtimes.get(k, 0)
        }
        # collect which files were expected to be modified without incurring any changes
        ok_modified_files = {
            _path_('.git/hooks/post-update'),
            'index.html',
        }
        ok_modified_files.add(_path_('.git/config'))
        ok_modified_files.update(
            {f
             for f in digests if f.startswith(_path_('.git/datalad/web'))})
        # it seems that with some recent git behavior has changed a bit
        # and index might get touched
        if _path_('.git/index') in modified_files:
            ok_modified_files.add(_path_('.git/index'))
        ok_(modified_files.issuperset(ok_modified_files))