mirror of
https://pagure.io/fedora-qa/os-autoinst-distri-fedora.git
synced 2024-11-09 17:04:21 +00:00
0b55fda6f5
Signed-off-by: Adam Williamson <awilliam@redhat.com>
182 lines
7.9 KiB
Python
182 lines
7.9 KiB
Python
# Copyright Red Hat
|
|
#
|
|
# This file is part of os-autoinst-distri-fedora.
|
|
#
|
|
# os-autoinst-distri-fedora is free software; you can redistribute it
|
|
# and/or modify it under the terms of the GNU General Public License
|
|
# as published by the Free Software Foundation, either version 2 of
|
|
# the License, or (at your option) any later version.
|
|
#
|
|
# This program is distributed in the hope that it will be useful,
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
# GNU General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU General Public License
|
|
# along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
#
|
|
# Author: Adam Williamson <awilliam@redhat.com>
|
|
|
|
# these are all kinda inappropriate for pytest patterns
|
|
# pylint: disable=no-init, protected-access, no-self-use, unused-argument
|
|
|
|
"""Tests for fifloader.py."""
|
|
|
|
# core imports
|
|
import json
|
|
import os
|
|
import tempfile
|
|
from unittest import mock
|
|
|
|
# third party imports
|
|
import jsonschema.exceptions
|
|
import pytest
|
|
|
|
# internal imports
|
|
import fifloader
|
|
|
|
DATAPATH = os.path.join(os.path.dirname(os.path.realpath(__file__)), 'data')
|
|
|
|
def _get_merged(input1='templates.fif.json', input2='templates-updates.fif.json'):
|
|
"""Convenience function as multiple tests need to do this."""
|
|
return fifloader.merge_inputs(
|
|
[os.path.join(DATAPATH, input1), os.path.join(DATAPATH, input2)])
|
|
|
|
def test_schema_validate():
|
|
"""Test for schema_validate."""
|
|
with open(os.path.join(DATAPATH, 'templates.fif.json'), 'r') as tempfh:
|
|
tempdata = json.load(tempfh)
|
|
with open(os.path.join(DATAPATH, 'templates-updates.fif.json'), 'r') as updfh:
|
|
updata = json.load(updfh)
|
|
assert fifloader.schema_validate(tempdata, fif=True, complete=True) is True
|
|
assert fifloader.schema_validate(tempdata, fif=True, complete=False) is True
|
|
assert fifloader.schema_validate(updata, fif=True, complete=False) is True
|
|
with pytest.raises(jsonschema.exceptions.ValidationError):
|
|
fifloader.schema_validate(updata, fif=True, complete=True)
|
|
with pytest.raises(jsonschema.exceptions.ValidationError):
|
|
fifloader.schema_validate(tempdata, fif=False, complete=True)
|
|
with pytest.raises(jsonschema.exceptions.ValidationError):
|
|
fifloader.schema_validate(tempdata, fif=False, complete=False)
|
|
# we test successful openQA validation later in test_run
|
|
|
|
# we test merging in both orders, because it can work in one order
|
|
# but be broken in the other
|
|
@pytest.mark.parametrize(
|
|
"input1, input2",
|
|
[
|
|
('templates.fif.json', 'templates-updates.fif.json'),
|
|
('templates-updates.fif.json', 'templates.fif.json'),
|
|
]
|
|
)
|
|
def test_merge_inputs(input1, input2):
|
|
"""Test for merge_inputs."""
|
|
(machines, products, profiles, testsuites, jobtemplates) = _get_merged(input1, input2)
|
|
# a few known attributes of the test data to ensure the merge worked
|
|
assert len(machines) == 2
|
|
assert len(products) == 4
|
|
assert len(profiles) == 4
|
|
assert not jobtemplates
|
|
# testsuite merging is the most complex feature
|
|
# len should be 3 as there is 1 unique suite in each input file,
|
|
# and one defined in both which should be merged
|
|
assert len(testsuites) == 3
|
|
# check the merged suite was merged correctly
|
|
# we should have the profiles from *both* input files...
|
|
assert len(testsuites['base_selinux']['profiles']) == 4
|
|
# and we should still have the settings (note, combining settings
|
|
# is not supported, the last-read settings dict is always used)
|
|
assert len(testsuites['base_selinux']['settings']) == 6
|
|
|
|
def test_generate_job_templates():
|
|
"""Test for generate_job_templates."""
|
|
(machines, products, profiles, testsuites, _) = _get_merged()
|
|
templates = fifloader.generate_job_templates(products, profiles, testsuites)
|
|
# we should get one template per profile in merged input
|
|
assert len(templates) == 8
|
|
for template in templates:
|
|
assert template['group_name'] in ['fedora', 'Fedora PowerPC', 'Fedora AArch64',
|
|
'Fedora Updates', 'Fedora PowerPC Updates',
|
|
'Fedora AArch64 Updates']
|
|
assert template['machine_name'] in list(machines.keys())
|
|
assert isinstance(template['prio'], int)
|
|
for item in ('arch', 'distri', 'flavor', 'version'):
|
|
assert item in template
|
|
assert template['test_suite_name'] in list(testsuites.keys())
|
|
|
|
def test_reverse_qol():
|
|
"""Test for reverse_qol."""
|
|
(machines, products, _, testsuites, _) = _get_merged()
|
|
(machines, products, testsuites) = fifloader.reverse_qol(machines, products, testsuites)
|
|
assert isinstance(machines, list)
|
|
assert isinstance(products, list)
|
|
assert isinstance(testsuites, list)
|
|
assert len(machines) == 2
|
|
assert len(products) == 4
|
|
assert len(testsuites) == 3
|
|
settlists = []
|
|
for datatype in (machines, products, testsuites):
|
|
for item in datatype:
|
|
# all items should have one of these
|
|
settlists.append(item['settings'])
|
|
# no items should have one of these
|
|
assert 'profiles' not in item
|
|
for settlist in settlists:
|
|
assert isinstance(settlist, list)
|
|
for setting in settlist:
|
|
assert list(setting.keys()) == ['key', 'value']
|
|
|
|
def test_parse_args():
|
|
"""Test for parse_args."""
|
|
args = fifloader.parse_args(['-l', '--host', 'https://openqa.example', '--clean', '--update',
|
|
'--loader', '/tmp/newloader', 'foo.json', 'bar.json'])
|
|
assert args.load is True
|
|
assert args.host == 'https://openqa.example'
|
|
assert args.clean is True
|
|
assert args.update is True
|
|
assert args.write is False
|
|
assert args.loader == '/tmp/newloader'
|
|
assert args.files == ['foo.json', 'bar.json']
|
|
args = fifloader.parse_args(['-l', '-w', 'foo.json', 'bar.json'])
|
|
assert args.load is True
|
|
assert not args.host
|
|
assert args.clean is False
|
|
assert args.update is False
|
|
assert args.write is True
|
|
assert args.filename == 'generated.json'
|
|
assert args.loader == '/usr/share/openqa/script/load_templates'
|
|
assert args.files == ['foo.json', 'bar.json']
|
|
args = fifloader.parse_args(['-w', '--filename', 'newout.json', 'foo.json'])
|
|
assert args.load is False
|
|
assert args.write is True
|
|
assert args.filename == 'newout.json'
|
|
assert args.files == ['foo.json']
|
|
|
|
@mock.patch('subprocess.run', autospec=True)
|
|
def test_run(fakerun):
|
|
"Test for run()."""
|
|
with pytest.raises(SystemExit, match=r".neither --write nor --load.*"):
|
|
fifloader.run(['--no-validate', 'foo.json'])
|
|
with pytest.raises(SystemExit) as excinfo:
|
|
fifloader.run(['-l'])
|
|
assert "arguments are required: files" in excinfo.message
|
|
with tempfile.NamedTemporaryFile() as tempfh:
|
|
# this will actually do everything and write out template data
|
|
# parsed from the test inputs to the temporary file
|
|
fifloader.run(['-w', '--filename', tempfh.name,
|
|
os.path.join(DATAPATH, 'templates.fif.json'),
|
|
os.path.join(DATAPATH, 'templates-updates.fif.json')])
|
|
written = json.load(tempfh)
|
|
# check written data matches upstream data schema
|
|
assert fifloader.schema_validate(written, fif=False, complete=True) is True
|
|
fifloader.run(['-l', '--loader', '/tmp/newloader', '--host',
|
|
'https://openqa.example', '--clean', '--update',
|
|
os.path.join(DATAPATH, 'templates.fif.json'),
|
|
os.path.join(DATAPATH, 'templates-updates.fif.json')])
|
|
assert fakerun.call_count == 1
|
|
assert fakerun.call_args[0][0] == ['/tmp/newloader', '--host', 'https://openqa.example',
|
|
'--clean', '--update', '-']
|
|
assert fakerun.call_args[1]['input'] == json.dumps(written)
|
|
assert fakerun.call_args[1]['text'] is True
|
|
|
|
# vim: set textwidth=100 ts=8 et sw=4:
|