# Licensed under the Apache License, Version 2.0 (the "License"); you may # not use this file except in compliance with the License. You may obtain # a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the # License for the specific language governing permissions and limitations # under the License. import argparse import csv import operator import requests import openstack_releases from openstack_releases import defaults from openstack_releases import deliverable from openstack_releases import schema from openstack_releases import series_status def get_stable_branch_id(series): """Retrieve the stable branch ID of the series. Returns the release-id if the series has such field, otherwise returns the series name. This is needed for the new stable branch naming style: stable/2023.1 (versus the old style: stable/zed). """ series_status_data = series_status.SeriesStatus.default() return series_status_data[series].release_id def main(): deliverable_schema = schema.Schema() parser = argparse.ArgumentParser() output_mode = parser.add_mutually_exclusive_group() output_mode.add_argument( '-v', '--verbose', action='store_true', default=False, help='show more than the deliverable name', ) output_mode.add_argument( '-r', '--repos', action='store_true', default=False, help='show the repository names not deliverable names', ) output_mode.add_argument( '-a', '--all-releases', action='store_true', default=False, help='show all of the releases for each deliverable', ) parser.add_argument( '--group-by', dest='group_key', default=None, choices=['team', 'type', 'model'], help='group output by the specified value', ) parser.add_argument( '--team', help='the name of the project team, such as "Nova" or "Oslo"', ) parser.add_argument( '--deliverable', help='the name of the deliverable, such as "nova" or "oslo.config"', ) parser.add_argument( '--series', default=defaults.RELEASE, help='the release series, such as "newton" or "ocata"', ) parser.add_argument( '--csvfile', help='Save results (same as when --verbose) to CSV file', ) parser.add_argument( '--show-dates', action='store_true', default=False, help='Show last release date (in verbose mode)', ) model = parser.add_mutually_exclusive_group() model.add_argument( '--model', help=('the release model, such as "cycle-with-rc"' ' or "independent"'), choices=sorted(deliverable_schema.release_models + ['independent']), ) model.add_argument( '--cycle-based', action='store_true', default=False, help='include all cycle-based deliverables', ) model.add_argument( '--cycle-based-no-trailing', action='store_true', default=False, help='include all cycle-based deliverables, except trailing ones', ) parser.add_argument( '--type', default=[], action='append', choices=sorted(deliverable_schema.release_types), help='deliverable type, such as "library" or "service"', ) parser.add_argument( '--deliverables-dir', default=openstack_releases.deliverable_dir, help='location of deliverable files', ) parser.add_argument( '--no-stable-branch', default=False, action='store_true', help='limit the list to deliverables without a stable branch', ) grp = parser.add_mutually_exclusive_group() grp.add_argument( '--unreleased', default=False, action='store_true', help='limit the list to deliverables not released in the cycle', ) grp.add_argument( '--unreleased-since', help=('limit the list to deliverables not released in the cycle ' 'since a given YYYY-MM-DD date'), ) grp.add_argument( '--missing-rc', action='store_true', help=('deliverables that do not have a release candidate, yet ' '(implies --model cycle-with-rc)'), ) grp.add_argument( '--is-eol', action='store_true', help='limit the list to deliverables EOL in the cycle', ) grp.add_argument( '--is-eom', action='store_true', help='limit the list to deliverables EOM in the cycle', ) grp.add_argument( '--missing-final', action='store_true', help='deliverables that have pre-releases but no final releases, yet', ) grp.add_argument( '--forced', action='store_true', help=('releases that have the "forced" flag applied ' '(implies --all-releases)'), ) args = parser.parse_args() series = args.series GET_REFS_API = 'https://opendev.org/api/v1/repos/{}/git/{}' GET_COMMIT_API = 'https://opendev.org/api/v1/repos/{}/git/commits/{}' if args.missing_rc: model = 'cycle-with-rc' version_ending = None elif args.missing_final: model = args.model version_ending = None else: model = args.model version_ending = None if args.unreleased_since: args.show_dates = True verbose_template = '{name:30} {team:20}' if not args.unreleased: verbose_template += ' {latest_release:12}' if args.show_dates: verbose_template += ' {last_release_date:11}' if len(args.type) != 1: verbose_template += ' {type:15}' if not args.model: verbose_template += ' {model:15}' if args.forced: args.all_releases = True csvfile = None if args.csvfile: csvfile = open(args.csvfile, 'w') fieldnames = ['name', 'latest_release', 'repo', 'hash', 'team', 'type', 'model'] writer = csv.DictWriter(csvfile, fieldnames=fieldnames) writer.writeheader() all_deliv = deliverable.Deliverables( root_dir=args.deliverables_dir, collapse_history=False, ) deliv_iter = list(all_deliv.get_deliverables(args.team, series)) if args.group_key: deliv_iter = sorted(deliv_iter, key=operator.attrgetter(args.group_key)) name_fmt = ' {}' else: name_fmt = '{}' cur_group = None for deliv in deliv_iter: if args.group_key: deliv_group = getattr(deliv, args.group_key) if args.deliverable and deliv.name != args.deliverable: continue if model and deliv.model != model: continue if args.cycle_based and not deliv.is_cycle_based: continue if args.cycle_based_no_trailing and (not deliv.is_cycle_based or deliv.type == 'trailing'): continue if args.type and deliv.type not in args.type: continue if args.no_stable_branch: if deliv.is_branchless: continue if deliv.name == 'release-test': continue if deliv.stable_branch_type is None: continue if deliv.get_branch_location('stable/' + get_stable_branch_id(series)) is not None: continue if args.unreleased and (deliv.is_released or not deliv.is_releasable): continue if args.is_eol and 'eol' not in deliv.latest_release: continue if args.is_eom and 'eom' not in deliv.latest_release: continue if version_ending and deliv.is_released: found = False for release in deliv.releases: if release.version.endswith(version_ending): found = True break if found: continue if args.missing_rc and deliv.is_released and 'rc' in deliv.latest_release: continue if args.missing_final and deliv.latest_release: if not ('rc' in deliv.latest_release or 'a' in deliv.latest_release or 'b' in deliv.latest_release): continue release_date = {} if (args.show_dates or args.unreleased_since) and deliv.is_released: if args.all_releases: versions = [a.version for a in deliv.releases] else: versions = [deliv.releases[-1].version] for ver in versions: ref = "refs/tags/{}".format(ver) api = GET_REFS_API.format(deliv.repos[0], ref) tagsjson = requests.get(api).json() # Gitea returns either a single tag object, or a list of # tag objects containing the provided string. So we need to # filter the list for the exact match. if isinstance(tagsjson, list): for release_tag in tagsjson: if release_tag['ref'] == ref: break else: release_tag = tagsjson release_sha = release_tag['object']['sha'] api = GET_COMMIT_API.format(deliv.repos[0], release_sha) release_commit = requests.get(api).json()['commit'] release_date[ver] = release_commit['author']['date'][0:10] if args.unreleased_since and deliv.is_released: if release_date[ver] >= args.unreleased_since: continue if csvfile: rel = (deliv.releases or [{}])[-1] for prj in rel.get('projects', [{}]): writer.writerow({ 'name': deliv.name, 'latest_release': rel.get('version', None), 'repo': prj.get('repo', None), 'hash': prj.get('hash', None), 'team': deliv.team, 'type': deliv.type, 'model': deliv.model, }) elif args.all_releases: for r in deliv.releases: if args.forced and not r.was_forced: continue print(verbose_template.format( name=deliv.name, latest_release=r.version, last_release_date=release_date.get(r.version, ''), team=deliv.team, type=deliv.type, model=deliv.model, )) elif args.verbose: print(verbose_template.format( name=deliv.name, latest_release=deliv.latest_release or '', last_release_date=release_date.get(deliv.latest_release, ''), team=deliv.team, type=deliv.type, model=deliv.model, )) elif args.repos: if args.group_key and cur_group != deliv_group: cur_group = deliv_group print(cur_group) for r in sorted(deliv.repos): print(name_fmt.format(r)) else: if args.group_key and cur_group != deliv_group: cur_group = deliv_group print(cur_group) print(name_fmt.format(deliv.name)) if csvfile: csvfile.close()