Stephan Kulow a26301bfa1 pkglistgen: Reset the tool between projects
So far I only cared for single projects, but never tested all staging
projects in a loop while developing, so I didn't notice the problem
2019-01-17 20:39:02 +01:00

738 lines
30 KiB
Python

from __future__ import print_function
import ToolBase
import glob
import logging
import os
import re
import solv
import shutil
import subprocess
import yaml
import sys
from lxml import etree as ET
from osc.core import checkout_package
from osc.core import http_GET
from osc.core import HTTPError
from osc.core import show_results_meta
from osc.core import Package
from osc.core import undelete_package
from osclib.core import attribute_value_load
from osclib.core import target_archs
from osclib.conf import str2bool
from osclib.core import repository_path_expand
from osclib.core import repository_arch_state
from osclib.cache_manager import CacheManager
try:
from urllib.parse import urlparse
except ImportError:
# python 2.x
from urlparse import urlparse
from pkglistgen import file_utils, solv_utils
from pkglistgen.group import Group
SCRIPT_PATH = os.path.dirname(os.path.realpath(__file__))
PRODUCT_SERVICE = '/usr/lib/obs/service/create_single_product'
class PkgListGen(ToolBase.ToolBase):
def __init__(self):
ToolBase.ToolBase.__init__(self)
self.logger = logging.getLogger(__name__)
self.reset()
def reset(self):
# package -> supportatus
self.packages = dict()
self.groups = dict()
self._supportstatus = None
self.input_dir = '.'
self.output_dir = '.'
self.lockjobs = dict()
self.ignore_broken = False
self.unwanted = set()
self.output = None
self.locales = set()
self.did_update = False
self.filtered_architectures = None
self.dry_run = False
self.all_architectures = None
def filter_architectures(self, architectures):
self.filtered_architectures = list(set(architectures) & set(self.all_architectures))
def _load_supportstatus(self):
# XXX
fn = os.path.join(self.input_dir, 'supportstatus.txt')
self._supportstatus = dict()
if os.path.exists(fn):
with open(fn, 'r') as fh:
for l in fh:
# pkg, status
a = l.rstrip().split(' ')
if len(a) > 1:
self._supportstatus[a[0]] = a[1]
def supportstatus(self, package):
if self._supportstatus is None:
self._load_supportstatus()
return self._supportstatus.get(package)
def _load_group_file(self, fn):
output = None
unwanted = None
with open(fn, 'r') as fh:
self.logger.debug('reading %s', fn)
for groupname, group in yaml.safe_load(fh).items():
if groupname == 'OUTPUT':
output = group
continue
if groupname == 'UNWANTED':
unwanted = set(group)
continue
g = Group(groupname, self)
g.parse_yml(group)
return output, unwanted
def load_all_groups(self):
for fn in glob.glob(os.path.join(self.input_dir, 'group*.yml')):
o, u = self._load_group_file(fn)
if o:
if self.output is not None:
raise Exception('OUTPUT defined multiple times')
self.output = o
if u:
self.unwanted |= u
# required to generate release spec files (only)
def write_group_stubs(self):
archs = ['*'] + self.all_architectures
for name in self.groups:
group = self.groups[name]
group.solved_packages = dict()
fn = '{}.group'.format(group.name)
with open(os.path.join(self.output_dir, fn), 'w') as fh:
for arch in archs:
x = group.toxml(arch, self.ignore_broken, None)
x = ET.tostring(x, pretty_print=True)
fh.write(x)
def write_all_groups(self):
self._check_supplements()
summary = dict()
archs = ['*'] + self.all_architectures
for name in self.groups:
group = self.groups[name]
if not group.solved:
continue
summary[name] = group.summary()
fn = '{}.group'.format(group.name)
with open(os.path.join(self.output_dir, fn), 'w') as fh:
comment = group.comment
for arch in archs:
x = group.toxml(arch, self.ignore_broken, comment)
# only comment first time
comment = None
x = ET.tostring(x, pretty_print=True)
x = re.sub(r'\s*<!-- reason:', ' <!-- reason:', x)
fh.write(x)
return summary
def solve_module(self, groupname, includes, excludes, use_recommends):
g = self.groups[groupname]
for i in includes:
g.inherit(self.groups[i])
g.solve(use_recommends)
for e in excludes:
g.ignore(self.groups[e])
def expand_repos(self, project, repo='standard'):
return repository_path_expand(self.apiurl, project, repo)
def _check_supplements(self):
tocheck = set()
tocheck_locales = set()
for arch in self.filtered_architectures:
pool = self._prepare_pool(arch)
sel = pool.Selection()
for s in pool.solvables_iter():
sel.add_raw(solv.Job.SOLVER_SOLVABLE, s.id)
for s in sel.solvables():
for dep in s.lookup_deparray(solv.SOLVABLE_SUPPLEMENTS):
for d in dep.str().split(' '):
if d.startswith('namespace:modalias') or d.startswith('namespace:filesystem'):
tocheck.add(s.name)
for l in self.locales:
i = pool.str2id('locale({})'.format(l))
for s in pool.whatprovides(i):
tocheck_locales.add(s.name)
all_grouped = set()
for g in self.groups.values():
if g.solved:
for arch in g.solved_packages.keys():
if g.solved_packages[arch]:
all_grouped.update(g.solved_packages[arch])
for p in tocheck - all_grouped:
self.logger.warn('package %s has supplements but is not grouped', p)
for p in tocheck_locales - all_grouped:
self.logger.warn('package %s provides supported locale but is not grouped', p)
def _prepare_pool(self, arch):
pool = solv.Pool()
pool.setarch(arch)
self.lockjobs[arch] = []
solvables = set()
for project, reponame in self.repos:
repo = pool.add_repo(project)
s = os.path.join(solv_utils.CACHEDIR, 'repo-{}-{}-{}.solv'.format(project, reponame, arch))
r = repo.add_solv(s)
if not r:
if not self.did_update:
raise Exception(
'failed to add repo {}/{}/{}. Need to run update first?'.format(project, reponame, arch))
continue
for solvable in repo.solvables_iter():
solvable.unset(solv.SOLVABLE_CONFLICTS)
solvable.unset(solv.SOLVABLE_OBSOLETES)
# only take the first solvable in the repo chain
if solvable.name in solvables:
self.lockjobs[arch].append(pool.Job(solv.Job.SOLVER_SOLVABLE | solv.Job.SOLVER_LOCK, solvable.id))
solvables.add(solvable.name)
pool.addfileprovides()
pool.createwhatprovides()
for l in self.locales:
pool.set_namespaceproviders(solv.NAMESPACE_LANGUAGE, pool.Dep(l), True)
return pool
# parse file and merge all groups
def _parse_unneeded(self, filename):
filename = os.path.join(self.input_dir, filename)
if not os.path.isfile(filename):
return set()
fh = open(filename, 'r')
self.logger.debug('reading %s', filename)
result = set()
for groupname, group in yaml.safe_load(fh).items():
result.update(group)
return result
# the unsorted group is special and will contain all the rest for
# the FTP tree. We filter it with unneeded though to create a
# unsorted.yml file for release manager review
def _collect_unsorted_packages(self, modules, unsorted):
uneeded_regexps = [re.compile(r)
for r in self._parse_unneeded('unneeded.yml')]
packages = dict()
if unsorted:
unsorted.solved_packages = dict()
unsorted.solved_packages['*'] = dict()
for arch in self.filtered_architectures:
pool = self._prepare_pool(arch)
pool.Selection()
archpacks = [s.name for s in pool.solvables_iter()]
# copy
filtered = list(archpacks)
for r in uneeded_regexps:
filtered = [p for p in filtered if not r.match(p)]
# convert to set
filtered = set(filtered) - self.unwanted
for g in modules:
if unsorted and g == unsorted:
continue
for a in ('*', arch):
filtered -= set(g.solved_packages[a].keys())
for package in filtered:
packages.setdefault(package, []).append(arch)
if unsorted:
archpacks = set(archpacks)
unsorted.solved_packages[arch] = dict()
for g in modules:
archpacks -= set(g.solved_packages[arch].keys())
archpacks -= set(g.solved_packages['*'].keys())
unsorted.solved_packages[arch] = dict()
for p in archpacks:
unsorted.solved_packages[arch][p] = None
if unsorted:
common = None
for arch in self.filtered_architectures:
if common is None:
common = set(unsorted.solved_packages[arch].keys())
continue
common &= set(unsorted.solved_packages[arch].keys())
for p in common:
unsorted.solved_packages['*'][p] = None
for arch in self.filtered_architectures:
del unsorted.solved_packages[arch][p]
with open(os.path.join(self.output_dir, 'unsorted.yml'), 'w') as fh:
fh.write('unsorted:\n')
for p in sorted(packages.keys()):
fh.write(' - ')
fh.write(p)
if len(packages[p]) != len(self.filtered_architectures):
fh.write(': [')
fh.write(','.join(sorted(packages[p])))
fh.write(']')
reason = self._find_reason(p, modules)
if reason:
fh.write(' # ' + reason)
fh.write(' \n')
# give a hint if the package is related to a group
def _find_reason(self, package, modules):
# go through the modules multiple times to find the "best"
for g in modules:
if package in g.recommends:
return 'recommended by ' + g.recommends[package]
for g in modules:
if package in g.suggested:
return 'suggested by ' + g.suggested[package]
for g in modules:
if package in g.develpkgs:
return 'devel package of ' + g.develpkgs[package]
return None
def update_repos(self, architectures):
# only there to parse the repos
bs_mirrorfull = os.path.join(SCRIPT_PATH, '..', 'bs_mirrorfull')
global_update = False
for project, repo in self.repos:
for arch in architectures:
# TODO: refactor to common function with repo_checker.py
d = os.path.join(solv_utils.CACHEDIR, project, repo, arch)
if not os.path.exists(d):
os.makedirs(d)
try:
# Fetch state before mirroring in-case it changes during download.
state = repository_arch_state(self.apiurl, project, repo, arch)
except HTTPError:
continue
# Would be preferable to include hash in name, but cumbersome to handle without
# reworking a fair bit since the state needs to be tracked.
solv_file = os.path.join(solv_utils.CACHEDIR, 'repo-{}-{}-{}.solv'.format(project, repo, arch))
solv_file_hash = '{}::{}'.format(solv_file, state)
if os.path.exists(solv_file) and os.path.exists(solv_file_hash):
# Solve file exists and hash unchanged, skip updating solv.
self.logger.debug('skipping solv generation for {} due to matching state {}'.format(
'/'.join([project, repo, arch]), state))
continue
# Either hash changed or new, so remove any old hash files.
file_utils.unlink_list(None, glob.glob(solv_file + '::*'))
global_update = True
self.logger.debug('updating %s', d)
args = [bs_mirrorfull]
args.append('--nodebug')
args.append('{}/public/build/{}/{}/{}'.format(self.apiurl, project, repo, arch))
args.append(d)
p = subprocess.Popen(args, stdout=subprocess.PIPE)
for line in p.stdout:
self.logger.info(line.rstrip())
files = [os.path.join(d, f)
for f in os.listdir(d) if f.endswith('.rpm')]
fh = open(solv_file, 'w')
p = subprocess.Popen(
['rpms2solv', '-m', '-', '-0'], stdin=subprocess.PIPE, stdout=fh)
p.communicate('\0'.join(files))
p.wait()
fh.close()
# Create hash file now that solv creation is complete.
open(solv_file_hash, 'a').close()
self.did_update = True
return global_update
def create_sle_weakremovers(self, target, oldprjs):
self.repos = []
for prj in list(oldprjs) + [target]:
self.repos += self.expand_repos(prj, 'standard')
self.update_repos(self.all_architectures)
drops = dict()
for arch in self.all_architectures:
pool = solv.Pool()
pool.setarch(arch)
sysrepo = None
for project, repo in self.repos:
self.logger.debug('processing %s/%s/%s', project, repo, arch)
fn = os.path.join(solv_utils.CACHEDIR, 'repo-{}-{}-{}.solv'.format(project, repo, arch))
r = pool.add_repo('/'.join([project, repo]))
r.add_solv(fn)
if project == target and repo == 'standard':
sysrepo = r
pool.createwhatprovides()
for s in pool.solvables_iter():
# we only want the old repos
if s.repo == sysrepo: continue
# ignore imported solvables. too dangerous
if s.arch != 'noarch' and s.arch != arch:
continue
haveit = False
for s2 in pool.whatprovides(s.nameid):
if s2.repo == sysrepo and s.nameid == s2.nameid:
haveit = True
if haveit:
continue
haveit = False
# check for already obsoleted packages
nevr = pool.rel2id(s.nameid, s.evrid, solv.REL_EQ)
for s2 in pool.whatmatchesdep(solv.SOLVABLE_OBSOLETES, nevr):
if s2.repo == sysrepo: continue
haveit = True
if haveit:
continue
drops.setdefault(s.name, {'repo': s.repo.name, 'archs': set()})
drops[s.name]['archs'].add(arch)
for project, repo in sorted(self.repos):
exclusives = dict()
print('#', project)
for name in sorted(drops.keys()):
#
if drops[name]['repo'] != '{}/{}'.format(project, repo):
#print(drops[name]['repo'], '!=', '{}/{}'.format(project, repo))
continue
if len(drops[name]['archs']) == len(self.all_architectures):
print('Provides: weakremover({})'.format(name))
else:
jarch = ' '.join(sorted(drops[name]['archs']))
exclusives.setdefault(jarch, []).append(name)
for arch in sorted(exclusives.keys()):
print('%ifarch {}'.format(arch))
for name in sorted(exclusives[arch]):
print('Provides: weakremover({})'.format(name))
print('%endif')
# TODO: no longer used, needs to be migrated
def create_droplist(self, output_dir, oldsolv):
drops = dict()
for arch in self.filtered_architectures:
for old in oldsolv:
self.logger.debug('%s: processing %s', arch, old)
pool = solv.Pool()
pool.setarch(arch)
for project, repo in self.repos:
fn = os.path.join(solv_utils.CACHEDIR, 'repo-{}-{}-{}.solv'.format(project, repo, arch))
r = pool.add_repo(project)
r.add_solv(fn)
sysrepo = pool.add_repo(os.path.basename(old).replace('.merged.solv', ''))
sysrepo.add_solv(old)
pool.createwhatprovides()
for s in sysrepo.solvables:
haveit = False
for s2 in pool.whatprovides(s.nameid):
if s2.repo == sysrepo or s.nameid != s2.nameid:
continue
haveit = True
if haveit:
continue
nevr = pool.rel2id(s.nameid, s.evrid, solv.REL_EQ)
for s2 in pool.whatmatchesdep(solv.SOLVABLE_OBSOLETES, nevr):
if s2.repo == sysrepo:
continue
haveit = True
if haveit:
continue
if s.name not in drops:
drops[s.name] = sysrepo.name
# mark it explicitly to avoid having 2 pools while GC is not run
del pool
ofh = sys.stdout
if output_dir:
name = os.path.join(output_dir, 'obsoletepackages.inc')
ofh = open(name, 'w')
for reponame in sorted(set(drops.values())):
print('<!-- %s -->' % reponame, file=ofh)
for p in sorted(drops):
if drops[p] != reponame:
continue
print(' <obsoletepackage>%s</obsoletepackage>' % p, file=ofh)
def solve_project(self, ignore_unresolvable=False, ignore_recommended=False, locale=None, locales_from=None):
self.load_all_groups()
if not self.output:
self.logger.error('OUTPUT not defined')
return
if ignore_unresolvable:
self.ignore_broken = True
global_use_recommends = not ignore_recommended
if locale:
for l in locale:
self.locales |= set(l.split(','))
if locales_from:
with open(os.path.join(self.input_dir, locales_from), 'r') as fh:
root = ET.parse(fh).getroot()
self.locales |= set([lang.text for lang in root.findall('.//linguas/language')])
modules = []
# the yml parser makes an array out of everything, so
# we loop a bit more than what we support
for group in self.output:
groupname = group.keys()[0]
settings = group[groupname]
if not settings: # e.g. unsorted
settings = {}
includes = settings.get('includes', [])
excludes = settings.get('excludes', [])
use_recommends = settings.get('recommends', global_use_recommends)
self.solve_module(groupname, includes, excludes, use_recommends)
g = self.groups[groupname]
g.conflicts = settings.get('conflicts', [])
g.default_support_status = settings.get('default-support', 'unsupported')
modules.append(g)
# not defined for openSUSE
overlap = self.groups.get('overlap')
for module in modules:
module.check_dups(modules, overlap)
module.collect_devel_packages()
module.filter_already_selected(modules)
if overlap:
ignores = [x.name for x in overlap.ignored]
self.solve_module(overlap.name, [], ignores, use_recommends=False)
overlapped = set(overlap.solved_packages['*'])
for arch in self.filtered_architectures:
overlapped |= set(overlap.solved_packages[arch])
for module in modules:
if module.name == 'overlap' or module in overlap.ignored:
continue
for arch in ['*'] + self.filtered_architectures:
for p in overlapped:
module.solved_packages[arch].pop(p, None)
self._collect_unsorted_packages(modules, self.groups.get('unsorted'))
return self.write_all_groups()
def strip_medium_from_staging(self, path):
# staging projects don't need source and debug medium - and the glibc source
# rpm conflicts between standard and bootstrap_copy repository causing the
# product builder to fail
medium = re.compile('name="(DEBUG|SOURCE)MEDIUM"')
for name in glob.glob(os.path.join(path, '*.kiwi')):
lines = open(name).readlines()
lines = [l for l in lines if not medium.search(l)]
open(name, 'w').writelines(lines)
def build_stub(self, destination, extension):
with open(os.path.join(destination, '.'.join(['stub', extension])), 'w+') as f:
f.write('# prevent building single {} files twice\n'.format(extension))
f.write('Name: stub\n')
f.write('Version: 0.0\n')
def commit_package(self, path):
if self.dry_run:
package = Package(path)
for i in package.get_diff():
print(''.join(i))
else:
# No proper API function to perform the same operation.
print(subprocess.check_output(
' '.join(['cd', path, '&&', 'osc', 'addremove']), shell=True))
package = Package(path)
package.commit(msg='Automatic update', skip_local_service_run=True)
def replace_product_version(self, product_file, product_version):
product_version = '<version>{}</version>'.format(product_version)
lines = open(product_file).readlines()
new_lines = []
for line in lines:
new_lines.append(line.replace('<version></version>', product_version))
open(product_file, 'w').write(''.join(new_lines))
def update_and_solve_target(self, api, target_project, target_config, main_repo,
project, scope, force, no_checkout,
only_release_packages, stop_after_solve, drop_list=False):
self.all_architectures = target_config.get('pkglistgen-archs').split(' ')
self.repos = self.expand_repos(project, main_repo)
print('[{}] {}/{}: update and solve'.format(scope, project, main_repo))
group = target_config.get('pkglistgen-group', '000package-groups')
product = target_config.get('pkglistgen-product', '000product')
release = target_config.get('pkglistgen-release', '000release-packages')
url = api.makeurl(['source', project])
packages = ET.parse(http_GET(url)).getroot()
if packages.find('entry[@name="{}"]'.format(product)) is None:
if not self.dry_run:
undelete_package(api.apiurl, project, product, 'revive')
# TODO disable build.
print('{} undeleted, skip dvd until next cycle'.format(product))
return
elif not force:
root = ET.fromstringlist(show_results_meta(api.apiurl, project, product,
repository=[main_repo], multibuild=True))
if len(root.xpath('result[@state="building"]')) or len(root.xpath('result[@state="dirty"]')):
print('{}/{} build in progress'.format(project, product))
return
checkout_list = [group, product, release]
if packages.find('entry[@name="{}"]'.format(release)) is None:
if not self.dry_run:
undelete_package(api.apiurl, project, release, 'revive')
print('{} undeleted, skip dvd until next cycle'.format(release))
return
# Cache dir specific to hostname and project.
host = urlparse(api.apiurl).hostname
cache_dir = CacheManager.directory('pkglistgen', host, project)
if not no_checkout:
if os.path.exists(cache_dir):
shutil.rmtree(cache_dir)
os.makedirs(cache_dir)
group_dir = os.path.join(cache_dir, group)
product_dir = os.path.join(cache_dir, product)
release_dir = os.path.join(cache_dir, release)
for package in checkout_list:
if no_checkout:
print('Skipping checkout of {}/{}'.format(project, package))
continue
checkout_package(api.apiurl, project, package, expand_link=True, prj_dir=cache_dir)
file_utils.unlink_all_except(release_dir)
if not only_release_packages:
file_utils.unlink_all_except(product_dir)
file_utils.copy_directory_contents(group_dir, product_dir,
['supportstatus.txt', 'groups.yml',
'reference-unsorted.yml', 'reference-summary.yml',
'package-groups.changes'])
file_utils.change_extension(product_dir, '.spec.in', '.spec')
file_utils.change_extension(product_dir, '.product.in', '.product')
self.input_dir = group_dir
self.output_dir = product_dir
print('-> do_update')
# make sure we only calculcate existant architectures
self.filter_architectures(target_archs(api.apiurl, project, main_repo))
self.update_repos(self.filtered_architectures)
if only_release_packages:
self.load_all_groups()
self.write_group_stubs()
else:
summary = self.solve_project(ignore_unresolvable=str2bool(target_config.get('pkglistgen-ignore-unresolvable')),
ignore_recommended=str2bool(target_config.get('pkglistgen-ignore-recommended')),
locale = target_config.get('pkglistgen-local'),
locales_from = target_config.get('pkglistgen-locales-from'))
if stop_after_solve:
return
delete_products = target_config.get('pkglistgen-delete-products', '').split(' ')
file_utils.unlink_list(product_dir, delete_products)
print('-> product service')
product_version = attribute_value_load(api.apiurl, project, 'ProductVersion')
if not product_version:
# for stagings the product version doesn't matter (I hope)
product_version = '1'
for product_file in glob.glob(os.path.join(product_dir, '*.product')):
self.replace_product_version(product_file, product_version)
print(subprocess.check_output(
[PRODUCT_SERVICE, product_file, product_dir, project]))
for delete_kiwi in target_config.get('pkglistgen-delete-kiwis-{}'.format(scope), '').split(' '):
delete_kiwis = glob.glob(os.path.join(product_dir, delete_kiwi))
file_utils.unlink_list(product_dir, delete_kiwis)
if scope == 'staging':
self.strip_medium_from_staging(product_dir)
spec_files = glob.glob(os.path.join(product_dir, '*.spec'))
file_utils.move_list(spec_files, release_dir)
inc_files = glob.glob(os.path.join(group_dir, '*.inc'))
file_utils.move_list(inc_files, release_dir)
file_utils.multibuild_from_glob(release_dir, '*.spec')
self.build_stub(release_dir, 'spec')
self.commit_package(release_dir)
if only_release_packages:
return
file_utils.multibuild_from_glob(product_dir, '*.kiwi')
self.build_stub(product_dir, 'kiwi')
self.commit_package(product_dir)
error_output = ''
reference_summary = os.path.join(group_dir, 'reference-summary.yml')
if os.path.isfile(reference_summary):
summary_file = os.path.join(product_dir, 'summary.yml')
with open(summary_file, 'w') as f:
f.write('# Summary of packages in groups')
for group in sorted(summary.keys()):
# the unsorted group should appear filtered by
# unneeded.yml - so we need the content of unsorted.yml
# not unsorted.group (this grew a little unnaturally)
if group == 'unsorted':
continue
f.write('\n' + group + ':\n')
for package in sorted(summary[group]):
f.write(' - ' + package + '\n')
try:
error_output += subprocess.check_output(['diff', '-u', reference_summary, summary_file])
except subprocess.CalledProcessError as e:
error_output += e.output
reference_unsorted = os.path.join(group_dir, 'reference-unsorted.yml')
unsorted_file = os.path.join(product_dir, 'unsorted.yml')
try:
error_output += subprocess.check_output(['diff', '-u', reference_unsorted, unsorted_file])
except subprocess.CalledProcessError as e:
error_output += e.output
if len(error_output) > 0:
self.logger.error('Difference in yml:\n' + error_output)
return True