There are 6 open security issues in bullseye.
commit 99bcc298e3598b7419f7686f30e920ea711d23e7
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Nov 8 17:59:05 2026 +0100
Point the quota classes patch to the merged review
The quota class sets shared-state fix was re-pushed upstream and
merged as review 1012547; review 1012435 stayed unmerged. Point the
Forwarded: header to the merged one.
commit 8b23923821d851ac2d73c9fcb836ef82cb9dc54b
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Nov 8 17:47:19 2026 +0100
Fix Forwarded: header of the two test-leak patches
fix-test-shared-fake-vcpu-mutation was merged upstream as review
1012545, and fix-test-monitor-jobstats-flag is open as review 1012544:
point the Forwarded: headers to them. Refresh both patches to remove
the hunk offsets so that they apply cleanly.
commit c1ef4227980067f4efa4b631ea383ca82f219bda
Author: Thomas Goirand <zigo@debian.org>
Date: Wed Oct 28 09:23:27 2026 +0100
Rework the guestfs forceTCG patch per upstream review feedback
Balazs Gibizer and Sean Mooney asked for the reset to live in the
fixture that pulls in the libvirt driver, so the flag cannot leak out
of any libvirt test, rather than in the guestfs vfs tests' setUp. The
quilt patch is regenerated from the amended upstream commit
1490bd210d: LibvirtFixture.setUp() now adds a cleanup calling
guestfs.force_tcg(False), and the vfs test setUp reset is gone.
Verified live: the flag set before LibvirtFixture.setUp() is False
after cleanUp().
commit 48588029232a6d249c808c8956bed0b375cdaae2
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 17:38:18 2026 +0100
Rework the compute mgr LOG patch per upstream review feedback
Balazs Gibizer asked for proper mocking on 1012436: instead of the
addCleanup(setattr) workaround, use the mock.patch.object(manager,
'LOG') decorator so manager.LOG is restored by the mock machinery
itself. The quilt patch is regenerated from the amended upstream
commit 064ac90190, validated with a poisoner pair (the reverts test
followed by test_finish_snapshot_based_resize_at_dest_spawn_fail_
graceful_rollback: the captured log assert fails before the fix, all
green after).
commit af1e064c768f4e0eff7cc9448e8753dc2184f5ce
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 17:32:11 2026 +0100
Rework the quota classes patch per upstream review feedback
Balazs Gibizer asked for the opposite approach on 1012435: instead of
rebinding fresh dicts in QuotaClassSetsTestV257.setUp(), initialize
quota_resources and filtered_quotas in the __init__ of
QuotaClassSetsTestV250 so that every instance (and every possibly
future child class) gets its own copy, and restore the original
V257.setUp mutations. The quilt patch is regenerated from the amended
upstream commit 48999d2643, validated with a poisoner pair (a V257
test followed by a V250 test: the documented jsonschema
ValidationError before the fix, all green after).
commit 2d27d72ea064b001bbcba8ba2b3f12a849ce7e17
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 17:18:22 2026 +0100
Fix the 2 remaining test failures of the -5 rebuild
Fixed leaking tests in nova/tests/unit/virt/libvirt/test_driver.py
(the shared fake_vcpumodel mutated by reference by a second test,
failing objects.test_instance.TestInstanceObject.test_create_with_extras)
and the leaked jobStats API disable flag
(JobInfo._have_job_stats flipped to False by any test raising
VIR_ERR_NO_SUPPORT, failing the live migration monitor job stats
tests), adding patches "Do not mutate the shared fake_vcpumodel in
the tests" and "Guard the live migration monitor tests against the
jobStats flag":
- fix-test-shared-fake-vcpu-mutation.patch
- fix-test-monitor-jobstats-flag.patch
Both leaks were validated with poisoner runs in a pristine 34.0.0
tree (the victim fails before the fix, passes after).
commit 00a195d6f0b0f3e94b5502156dd05cbf05b80a3e
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 17:04:42 2026 +0100
Fix the 4 remaining test failures of the -5 rebuild
Fixed leaking tests in nova/tests/unit/virt/libvirt/test_driver.py
(the shared fake_vcpumodel mutated by reference), leaking config
dynamic opts registration in nova/tests/unit/conf/ (mdev_nvidia-11
and neutron_physnet_foo left in the global CONF) and the leaked
guestfs forceTCG module-level flag, all of which fail other tests
depending on the worker partitioning of the dynamic stestr scheduler,
adding patches "Do not mutate the shared fake_vcpumodel in a test",
"Use a private ConfigOpts in the conf dynamic-opts tests" and "Reset
the guestfs forceTCG flag in the vfs tests setUp":
- fix-test-shared-fake-vcpu-mutation.patch
- fix-test-conf-dynamic-opts-hermetic.patch
- fix-test-guestfs-forcetcg-reset.patch
Each leak was reproduced with a poisoner pair run in a pristine
34.0.0 tree (the pair fails before the fix, passes after), and the
leakers that could not be pinned to a specific test in the build
environment are covered by making the victims hermetic.
commit 737b8b88d37e40e2e8db1edacdc00c73e4f8e4ea
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 16:24:11 2026 +0100
Blacklist eventlet-hostile tests instead of carrying fixes
Upstream rejected the greenthread-mode fixes for the delayed task
scheduler, the manager graceful shutdown and the websocketproxy hub
replacement, because eventlet is being removed. Drop the patches and
blacklist the tests that spawn eventlet-hostile native threads
(virt.libvirt.test_host, test_thread_pool_factory, the conductor
ConductorTask* classes, test_service.ServiceTestCase, the service API
and policy tests) as well as the whole console.test_websocketproxy
module, which replaces the eventlet hub on every websocket connection.
Validated by replaying the 32 worker histories of the wedged -5 build
in a pristine 34.0.0 tree: without the blacklist 659 tests fail, with
it all 32 workers complete with only the pre-existing replay
artifacts (43, versus 49 with the rejected fixes applied), and no
native thread of the three leaking classes spawns anymore.
commit 174e4694e1bdbd4bec34cf60b4f2799736d76484
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:47:35 2026 +0100
Stamp the CastAsCallFixture patch with its Gerrit URL
commit 02ec9732c878444cd6c2de5be6733d499063a1e4
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:46:54 2026 +0100
Stamp the compute mgr LOG patch with its Gerrit URL
commit 46dfa6e87fbe70d59967b755707d7ace225bc992
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:46:06 2026 +0100
Stamp the quota classes patch with its Gerrit URL
commit 55440509a3c3e7108e3e2eab96d5301cf26917a4
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:45:12 2026 +0100
Stamp the resource tracker fixture patch with its Gerrit URL
commit 5dc04eb57b177235e96bcc1d335f0244869cf131
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:43:56 2026 +0100
Stamp the hypervisors dict mutation patch with its Gerrit URL
commit 9ae4f7a50827d63167a0971f726256e694aeb84d
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:42:48 2026 +0100
Stamp the notifications time override patch with its Gerrit URL
commit 7813757ace8f07651d4a1a4a85fa779ffd47a881
Author: Thomas Goirand <zigo@debian.org>
Date: Tue Oct 27 09:27:31 2026 +0100
Fix CastAsCallFixture leaving a phantom cast entry
CastAsCallFixture._stub_out() stubs
oslo_messaging.rpc.client._CallContext.cast on every prepared
context. The cast attribute is inherited from _BaseCallContext and
not defined in _CallContext's own __dict__, so the monkeypatch
restore writes the inherited value into the subclass __dict__,
leaving a phantom _CallContext.cast entry behind after the test.
Later tests in the same process patching
oslo_messaging.rpc.client._BaseCallContext.cast (e.g.
ComputeAPIUnitTestCase.test_allow_share/.test_deny_share) then have
their mock shadowed by the phantom entry: cctxt.cast resolves
through _CallContext first and the _BaseCallContext.cast mock never
sees the call, failing with 'AssertionError: expected call not
found. Actual: not called.' This made nova FTBFS nondeterministically
under stestr --dynamic scheduling, where a fixture user and the
victim share a worker process.
Resolve the class that actually defines the attribute in its own
__dict__ (MRO walk) and stub that one, so the restore is
artifact-free for both stubs. Behavior of the stubs is unchanged.
Assisted-By: Infomaniak Euria-Code
commit e8b94561362795bf8cb95bfe643dd42c4ff489d5
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Oct 25 01:25:56 2026 +0200
Restore manager.LOG after the reverts_task_state test
Fixed leaking tests in
nova/tests/unit/compute/test_compute_mgr.py that leads to FTBFS,
adding patch "Restore manager.LOG after the reverts_task_state
test":
- fix-test-compute-mgr-log-rebind.patch
commit de3a53d9b96cda5b6a703751f01ae224f53d0bc0
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Oct 25 01:24:19 2026 +0200
Stop quota classes tests mutating shared filtered quotas
Fixed leaking tests in
nova/tests/unit/api/openstack/compute/test_quota_classes.py that
leads to FTBFS, adding patch "Stop QuotaClassSetsTestV257.setUp()
from mutating shared state":
- fix-test-quota-classes-filtered-quotas-mutation.patch
commit cabce881bf05a8eb3303123e0b790aefea0379ca
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Oct 25 01:18:33 2026 +0200
Restore resource tracker test fixtures after each test
Fixed leaking tests in
nova/tests/unit/compute/test_resource_tracker.py that leads to
FTBFS, adding patch "Restore the resource tracker test fixtures
after each test":
- fix-test-resource-tracker-fixture-mutation.patch
commit bfd0a728ea9c59878b93972ba5db20cc65f0007e
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Oct 25 01:08:43 2026 +0200
Fix hypervisors tests mutating a shared class dict
Fixed leaking tests in
nova/tests/unit/api/openstack/compute/test_hypervisors.py that
leads to FTBFS, adding patch "Stop HypervisorsTestV253.setUp()
from mutating a shared dict":
- fix-test-hypervisors-class-dict-mutation.patch
commit 39e3e4e9f9ddf7b875f627af8e1aab5506e36304
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Oct 25 01:07:16 2026 +0200
Fix notifications tests leaking their time override
Fixed leaking tests in nova/tests/unit/test_notifications.py
that leads to FTBFS, adding patch "Clear the leaked time
override in the notifications tests":
- fix-test-notifications-time-override-leak.patch
commit a9c0c449ca8c84c36bf34eeed533f2fe3bca059f
Author: Thomas Goirand <zigo@debian.org>
Date: Sun Oct 25 00:38:02 2026 +0200
--dynamic-debug
commit f500a4be0ec2fad101796d7b04a8dfbbc6bb0bac
Author: Thomas Goirand <zigo@debian.org>
Date: Sat Oct 24 11:16:32 2026 +0200
Switch to --dynamic mode for stestr.
commit a2da7a48674a5fade0071e56ad4d19d5584fbc23
Author: Thomas Goirand <zigo@debian.org>
Date: Fri Oct 16 12:00:10 2026 +0200
Releasing to unstable.
There is 1 open security issue in trixie.
You can find information about how to handle this issue in the security team's documentation.
Among the 10 debian patches available in version 2:34.0.0-4 of the package, we noticed the following issues:
Automatic checks made by the Debian l10n team found some issues with the translations contained in this package. You should check the l10n status report for more information.
Issues can be things such as missing translations, problematic translated strings, outdated PO files, unknown languages, etc.
![[bug history graph]](https://qa.debian.org/data/bts/graphs/n/nova.png){kind=link}