def check_mirror_url(mirror_url):
url = mirror_url.url + 'lastsync'
logger.info("checking URL %s", url)
log = MirrorLog(url=mirror_url, check_time=utc_now())
try:
start = time.time()
result = urllib2.urlopen(url, timeout=10)
data = result.read()
result.close()
end = time.time()
# lastsync should be an epoch value created by us
parsed_time = None
try:
parsed_time = datetime.utcfromtimestamp(int(data))
parsed_time = parsed_time.replace(tzinfo=utc)
except ValueError:
# it is bad news to try logging the lastsync value;
# sometimes we get a crazy-encoded web page.
pass
log.last_sync = parsed_time
# if we couldn't parse a time, this is a failure
if parsed_time is None:
log.error = "Could not parse time from lastsync"
log.is_success = False
log.duration = end - start
logger.debug("success: %s, %.2f", url, log.duration)
except urllib2.HTTPError, e:
if e.code == 404:
# we have a duration, just not a success
end = time.time()
log.duration = end - start
log.is_success = False
log.error = str(e)
logger.debug("failed: %s, %s", url, log.error)
def populate_files(dbpkg, repopkg, force=False):
if not force:
if not pkg_same_version(repopkg, dbpkg):
logger.info("DB version (%s) didn't match repo version "
"(%s) for package %s, skipping file list addition",
dbpkg.full_version, repopkg.full_version, dbpkg.pkgname)
return
if not dbpkg.files_last_update or not dbpkg.last_update:
pass
elif dbpkg.files_last_update > dbpkg.last_update:
return
# only delete files if we are reading a DB that contains them
if repopkg.has_files:
delete_pkg_files(dbpkg)
logger.info("adding %d files for package %s",
len(repopkg.files), dbpkg.pkgname)
pkg_files = []
for f in repopkg.files:
dirname, filename = f.rsplit('/', 1)
if filename == '':
filename = None
pkgfile = PackageFile(pkg=dbpkg,
is_directory=(filename is None),
directory=dirname + '/',
filename=filename)
pkg_files.append(pkgfile)
PackageFile.objects.bulk_create(pkg_files)
dbpkg.files_last_update = utc_now()
dbpkg.save()
def get_mirror_url_for_download(cutoff=default_cutoff):
'''Find a good mirror URL to use for package downloads. If we have mirror
status data available, it is used to determine a good choice by looking at
the last batch of status rows.'''
cutoff_time = utc_now() - cutoff
status_data = MirrorLog.objects.filter(
check_time__gte=cutoff_time).aggregate(
Max('check_time'), Max('last_sync'))
if status_data['check_time__max'] is not None:
min_check_time = status_data['check_time__max'] - timedelta(minutes=5)
min_sync_time = status_data['last_sync__max'] - timedelta(minutes=30)
best_logs = MirrorLog.objects.filter(is_success=True,
check_time__gte=min_check_time, last_sync__gte=min_sync_time,
url__mirror__public=True, url__mirror__active=True,
url__protocol__protocol__iexact='HTTP').order_by(
'duration')[:1]
if best_logs:
return MirrorUrl.objects.get(id=best_logs[0].url_id)
mirror_urls = MirrorUrl.objects.filter(
mirror__public=True, mirror__active=True,
protocol__protocol__iexact='HTTP')
# look first for an 'Any' URL, then fall back to any HTTP URL
filtered_urls = mirror_urls.filter(mirror__country='Any')[:1]
if not filtered_urls:
filtered_urls = mirror_urls[:1]
if not filtered_urls:
return None
return filtered_urls[0]
def set_news_fields(sender, **kwargs):
news = kwargs['instance']
now = utc_now()
news.last_modified = now
if not news.postdate:
news.postdate = now
# http://diveintomark.org/archives/2004/05/28/howto-atom-id
news.guid = 'tag:%s,%s:%s' % (Site.objects.get_current(),
now.strftime('%Y-%m-%d'), news.get_absolute_url())
def perform_updates():
now = utc_now()
pkgs.update(flag_date=now)
# store our flag request
flag_request = FlagRequest(created=now,
user_email=email, message=message,
ip_address=ip_addr, pkgbase=pkg.pkgbase,
version=version, repo=pkg.repo,
num_packages=len(flagged_pkgs))
if request.user.is_authenticated():
flag_request.user = request.user
flag_request.save()
def get_mirror_statuses(cutoff=default_cutoff):
cutoff_time = utc_now() - cutoff
protocols = list(MirrorProtocol.objects.filter(is_download=True))
# I swear, this actually has decent performance...
urls = MirrorUrl.objects.select_related('mirror', 'protocol').filter(
mirror__active=True, mirror__public=True,
protocol__in=protocols,
logs__check_time__gte=cutoff_time).annotate(
check_count=Count('logs'),
success_count=Count('logs__duration'),
last_sync=Max('logs__last_sync'),
last_check=Max('logs__check_time'),
duration_avg=Avg('logs__duration'),
duration_stddev=StdDev('logs__duration')
).order_by('-last_sync', '-duration_avg')
# The Django ORM makes it really hard to get actual average delay in the
# above query, so run a seperate query for it and we will process the
# results here.
times = MirrorLog.objects.filter(is_success=True, last_sync__isnull=False,
check_time__gte=cutoff_time)
delays = {}
for log in times:
delay = log.check_time - log.last_sync
delays.setdefault(log.url_id, []).append(delay)
if urls:
last_check = max([u.last_check for u in urls])
num_checks = max([u.check_count for u in urls])
check_info = MirrorLog.objects.filter(
check_time__gte=cutoff_time).aggregate(
mn=Min('check_time'), mx=Max('check_time'))
if num_checks > 1:
check_frequency = (check_info['mx'] - check_info['mn']) \
/ (num_checks - 1)
else:
check_frequency = None
else:
last_check = None
num_checks = 0
check_frequency = None
for url in urls:
annotate_url(url, delays)
return {
'cutoff': cutoff,
'last_check': last_check,
'num_checks': num_checks,
'check_frequency': check_frequency,
'urls': urls,
}
def get_mirror_errors(cutoff=default_cutoff):
cutoff_time = utc_now() - cutoff
errors = MirrorLog.objects.filter(
is_success=False, check_time__gte=cutoff_time,
url__mirror__active=True, url__mirror__public=True).values(
'url__url', 'url__country', 'url__protocol__protocol',
'url__mirror__country', 'error').annotate(
error_count=Count('error'), last_occurred=Max('check_time')
).order_by('-last_occurred', '-error_count')
errors = list(errors)
for err in errors:
err['country'] = err['url__country'] or err['url__mirror__country']
return errors
def clock(request):
devs = User.objects.filter(is_active=True).order_by(
'first_name', 'last_name').select_related('userprofile')
now = utc_now()
page_dict = {
'developers': devs,
'utc_now': now,
}
response = direct_to_template(request, 'devel/clock.html', page_dict)
if not response.has_header('Expires'):
expire_time = now.replace(second=0, microsecond=0)
expire_time += timedelta(minutes=1)
expire_time = time.mktime(expire_time.timetuple())
response['Expires'] = http_date(expire_time)
return response
def handle(self, *args, **options):
parser = IsoListParser()
isonames = Iso.objects.values_list('name', flat=True)
active_isos = parser.parse(settings.ISO_LIST_URL)
for iso in active_isos:
# create any names that don't already exist
if iso not in isonames:
new = Iso(name=iso, active=True)
new.save()
# update those that do if they were marked inactive
else:
existing = Iso.objects.get(name=iso)
if not existing.active:
existing.active = True
existing.removed = None
existing.save()
now = utc_now()
# and then mark all other names as no longer active
Iso.objects.filter(active=True).exclude(name__in=active_isos).update(
active=False, removed=now)
def signoff_package(request, name, repo, arch, revoke=False):
packages = get_list_or_404(Package, pkgbase=name,
arch__name=arch, repo__name__iexact=repo, repo__testing=True)
package = packages[0]
spec = SignoffSpecification.objects.get_or_default_from_package(package)
if revoke:
try:
signoff = Signoff.objects.get_from_package(
package, request.user, False)
except Signoff.DoesNotExist:
raise Http404
signoff.revoked = utc_now()
signoff.save()
created = False
else:
# ensure we should even be accepting signoffs
if spec.known_bad or not spec.enabled:
return render(request, '403.html', status=403)
signoff, created = Signoff.objects.get_or_create_from_package(
package, request.user)
all_signoffs = Signoff.objects.for_package(package)
if request.is_ajax():
data = {
'created': created,
'revoked': bool(signoff.revoked),
'approved': approved_by_signoffs(all_signoffs, spec),
'required': spec.required,
'enabled': spec.enabled,
'known_bad': spec.known_bad,
'user': str(request.user),
}
return HttpResponse(simplejson.dumps(data, ensure_ascii=False),
mimetype='application/json')
return redirect('package-signoffs')
def report(request, report_name, username=None):
title = 'Developer Report'
packages = Package.objects.normal()
names = attrs = user = None
if username:
user = get_object_or_404(User, username=username, is_active=True)
maintained = PackageRelation.objects.filter(user=user,
type=PackageRelation.MAINTAINER).values('pkgbase')
packages = packages.filter(pkgbase__in=maintained)
maints = User.objects.filter(id__in=PackageRelation.objects.filter(
type=PackageRelation.MAINTAINER).values('user'))
if report_name == 'old':
title = 'Packages last built more than one year ago'
cutoff = utc_now() - timedelta(days=365)
packages = packages.filter(
build_date__lt=cutoff).order_by('build_date')
elif report_name == 'long-out-of-date':
title = 'Packages marked out-of-date more than 90 days ago'
cutoff = utc_now() - timedelta(days=90)
packages = packages.filter(
flag_date__lt=cutoff).order_by('flag_date')
elif report_name == 'big':
title = 'Packages with compressed size > 50 MiB'
cutoff = 50 * 1024 * 1024
packages = packages.filter(
compressed_size__gte=cutoff).order_by('-compressed_size')
names = [ 'Compressed Size', 'Installed Size' ]
attrs = [ 'compressed_size_pretty', 'installed_size_pretty' ]
# Format the compressed and installed sizes with MB/GB/etc suffixes
for package in packages:
package.compressed_size_pretty = filesizeformat(
package.compressed_size)
package.installed_size_pretty = filesizeformat(
package.installed_size)
elif report_name == 'badcompression':
title = 'Packages that have little need for compression'
cutoff = 0.90 * F('installed_size')
packages = packages.filter(compressed_size__gt=0, installed_size__gt=0,
compressed_size__gte=cutoff).order_by('-compressed_size')
names = [ 'Compressed Size', 'Installed Size', 'Ratio', 'Type' ]
attrs = [ 'compressed_size_pretty', 'installed_size_pretty',
'ratio', 'compress_type' ]
# Format the compressed and installed sizes with MB/GB/etc suffixes
for package in packages:
package.compressed_size_pretty = filesizeformat(
package.compressed_size)
package.installed_size_pretty = filesizeformat(
package.installed_size)
ratio = package.compressed_size / float(package.installed_size)
package.ratio = '%.2f' % ratio
package.compress_type = package.filename.split('.')[-1]
elif report_name == 'uncompressed-man':
title = 'Packages with uncompressed manpages'
# checking for all '.0'...'.9' + '.n' extensions
bad_files = PackageFile.objects.filter(is_directory=False,
directory__contains='/man/',
filename__regex=r'\.[0-9n]').exclude(
filename__endswith='.gz').exclude(
filename__endswith='.xz').exclude(
filename__endswith='.bz2').exclude(
filename__endswith='.html')
if username:
pkg_ids = set(packages.values_list('id', flat=True))
bad_files = bad_files.filter(pkg__in=pkg_ids)
bad_files = bad_files.values_list('pkg_id', flat=True).distinct()
packages = packages.filter(id__in=set(bad_files))
elif report_name == 'uncompressed-info':
title = 'Packages with uncompressed infopages'
# we don't worry about looking for '*.info-1', etc., given that an
# uncompressed root page probably exists in the package anyway
bad_files = PackageFile.objects.filter(is_directory=False,
directory__endswith='/info/', filename__endswith='.info')
if username:
pkg_ids = set(packages.values_list('id', flat=True))
bad_files = bad_files.filter(pkg__in=pkg_ids)
bad_files = bad_files.values_list('pkg_id', flat=True).distinct()
packages = packages.filter(id__in=set(bad_files))
elif report_name == 'unneeded-orphans':
title = 'Orphan packages required by no other packages'
owned = PackageRelation.objects.all().values('pkgbase')
required = PackageDepend.objects.all().values('depname')
# The two separate calls to exclude is required to do the right thing
packages = packages.exclude(pkgbase__in=owned).exclude(
pkgname__in=required)
elif report_name == 'mismatched-signature':
title = 'Packages with mismatched signatures'
names = [ 'Signature Date', 'Signed By', 'Packager' ]
attrs = [ 'sig_date', 'sig_by', 'packager' ]
cutoff = timedelta(hours=24)
finder = UserFinder()
filtered = []
packages = packages.filter(pgp_signature__isnull=False)
for package in packages:
sig_date = package.signature.datetime.replace(tzinfo=pytz.utc)
package.sig_date = sig_date.date()
key_id = package.signature.key_id
signer = finder.find_by_pgp_key(key_id)
package.sig_by = signer or key_id
if signer is None or signer.id != package.packager_id:
filtered.append(package)
elif sig_date > package.build_date + cutoff:
filtered.append(package)
packages = filtered
else:
raise Http404
context = {
'all_maintainers': maints,
'title': title,
'maintainer': user,
'packages': packages,
'column_names': names,
'column_attrs': attrs,
}
return direct_to_template(request, 'devel/packages.html', context)
def db_update(archname, reponame, pkgs, force=False):
"""
Parses a list of packages and updates the packages database accordingly.
"""
logger.info('Updating %s (%s)', reponame, archname)
dbpkgs = update_common(archname, reponame, pkgs, True)
repository = Repo.objects.get(name__iexact=reponame)
architecture = Arch.objects.get(name__iexact=archname)
# This makes our inner loop where we find packages by name *way* more
# efficient by not having to go to the database for each package to
# SELECT them by name.
dbdict = dict((dbpkg.pkgname, dbpkg) for dbpkg in dbpkgs)
dbset = set(dbdict.keys())
syncset = set([pkg.name for pkg in pkgs])
in_sync_not_db = syncset - dbset
logger.info("%d packages in sync not db", len(in_sync_not_db))
# packages in syncdb and not in database (add to database)
for pkg in (pkg for pkg in pkgs if pkg.name in in_sync_not_db):
logger.info("Adding package %s", pkg.name)
dbpkg = Package(pkgname=pkg.name, arch=architecture, repo=repository)
try:
with transaction.commit_on_success():
populate_pkg(dbpkg, pkg, timestamp=utc_now())
except IntegrityError:
logger.warning("Could not add package %s; "
"not fatal if another thread beat us to it.",
pkg.name, exc_info=True)
# packages in database and not in syncdb (remove from database)
for pkgname in (dbset - syncset):
logger.info("Removing package %s", pkgname)
dbpkg = dbdict[pkgname]
with transaction.commit_on_success():
# no race condition here as long as simultaneous threads both
# issue deletes; second delete will be a no-op
delete_pkg_files(dbpkg)
dbpkg.delete()
# packages in both database and in syncdb (update in database)
pkg_in_both = syncset & dbset
for pkg in (x for x in pkgs if x.name in pkg_in_both):
logger.debug("Checking package %s", pkg.name)
dbpkg = dbdict[pkg.name]
timestamp = None
# for a force, we don't want to update the timestamp.
# for a non-force, we don't want to do anything at all.
if not force and pkg_same_version(pkg, dbpkg):
continue
elif not force:
timestamp = utc_now()
# The odd select_for_update song and dance here are to ensure
# simultaneous updates don't happen on a package, causing
# files/depends/all related items to be double-imported.
with transaction.commit_on_success():
dbpkg = Package.objects.select_for_update().get(id=dbpkg.id)
if not force and pkg_same_version(pkg, dbpkg):
logger.debug("Package %s was already updated", pkg.name)
continue
logger.info("Updating package %s", pkg.name)
populate_pkg(dbpkg, pkg, force=force, timestamp=timestamp)
logger.info('Finished updating arch: %s', archname)
def generate_report(email, repo_name):
repo = Repo.objects.get(name__iexact=repo_name)
# Collect all existing signoffs for these packages
signoff_groups = sorted(get_signoff_groups([repo]),
key=attrgetter('target_repo', 'arch', 'pkgbase'))
disabled = []
bad = []
complete = []
incomplete = []
new = []
old = []
new_hours = 24
old_days = 14
now = utc_now()
new_cutoff = now - timedelta(hours=new_hours)
old_cutoff = now - timedelta(days=old_days)
if len(signoff_groups) == 0:
# no need to send an email at all
return
for group in signoff_groups:
spec = group.specification
if spec.known_bad:
bad.append(group)
elif not spec.enabled:
disabled.append(group)
elif group.approved():
complete.append(group)
else:
incomplete.append(group)
if group.package.last_update > new_cutoff:
new.append(group)
if group.package.last_update < old_cutoff:
old.append(group)
old.sort(key=attrgetter('last_update'))
proto = 'https'
domain = Site.objects.get_current().domain
signoffs_url = '%s://%s%s' % (proto, domain, reverse('package-signoffs'))
# and the fun bit
Leader = namedtuple('Leader', ['user', 'count'])
leaders = Signoff.objects.filter(created__gt=new_cutoff,
revoked__isnull=True).values_list('user').annotate(
signoff_count=Count('pk')).order_by('-signoff_count')[:5]
users = User.objects.in_bulk([l[0] for l in leaders])
leaders = (Leader(users[l[0]], l[1]) for l in leaders)
subject = 'Signoff report for [%s]' % repo.name.lower()
t = loader.get_template('packages/signoff_report.txt')
c = Context({
'repo': repo,
'signoffs_url': signoffs_url,
'disabled': disabled,
'bad': bad,
'all': signoff_groups,
'incomplete': incomplete,
'complete': complete,
'new': new,
'new_hours': new_hours,
'old': old,
'old_days': old_days,
'leaders': leaders,
})
from_addr = 'Arch Website Notification <*****@*****.**>'
send_mail(subject, t.render(c), from_addr, [email])
