more relnotes updates

This commit is contained in:
Mike McLean 2020-07-27 12:19:32 -04:00 committed by Yu Ming Zhu
parent 20b2a01879
commit 59e6866940

View file

@ -17,7 +17,8 @@ For details on migrating see :doc:`../migrations/migrating_to_1.22`
Security Fixes
--------------
`None`
None
Client Changes
@ -27,14 +28,7 @@ Client Changes
| PR: https://pagure.io/koji/pull-request/2255
``mock-config`` command honors 'rpm.macro.*' options in tag's extra config now.
**Unify error messages in CLI**
| PR: https://pagure.io/koji/pull-request/2044
This is a technical dept enhancement. Now we are calling ``error()`` instead of
``print()`` and ``return 1``.
The ``mock-config`` command honors 'rpm.macro.*' options in tag's extra config now.
**--ca option has been deprecated**
@ -59,20 +53,24 @@ In some CLI commands we used ``active_session()`` which will try its best to
login, but it is not necessary. Now, we only ensure the connection without
authentication.
**Hide local --debug options for some commands**
**Unify --debug options**
| PR: https://pagure.io/koji/pull-request/2085
Following commands use the global ``--debug`` option for their own debug purpose
now:
The cli accepts a global ``--debug`` option before the command name.
Some commands accepted a separate ``--debug`` option local to the command,
which was confusing.
Now these commands take their cue from the global option.
The local option is still accepted for backwards compatibility, though
it has been hidden in the help output.
The following commands were affected:
* ``prune-sigs``
* ``list-signed``
* ``list-tag-history``
* ``list-history``
The local ``--debug`` options are still usable as we just hide them in helps.
**New option --wait for download-task**
| PR: https://pagure.io/koji/pull-request/2346
@ -84,8 +82,9 @@ finished as the same as the behavior of ``build`` command.
| PR: https://pagure.io/koji/pull-request/2347
Although there's a ``--wait`` option for ``image-build-indirection``, it was not
working until it's fixed by this PR.
Previously, the ``image-build-indirection`` command accepted the ``--wait``
option, but did not honor it.
This oversight has been fixed.
**Fix event option handling in clone-tag**
@ -117,6 +116,7 @@ malfunctions, like cache, filesystem, network, etc. Now, in
* check the rpm header is valid if the file is an RPM
* do 3 times retries if it fails
API Changes
-----------
@ -135,15 +135,19 @@ argument is not expected.
This host API will be finally removed in 1.23
**query_buildroots have to return ASAP**
**Optimizations to the listBuildroots call**
| PR: https://pagure.io/koji/pull-request/2299
| PR: https://pagure.io/koji/pull-request/2303
| PR: https://pagure.io/koji/pull-request/2301
For the optimization purpose, ``query_buildroots()`` can return earlier if there
is no candidate buildroot. This function equals ``listBuildroots`` API and is
used by ``getBuildroot``.
For the optimization purpose, the ``listBuildroots`` API call avoids
unnecessary checks when the return will be empty.
Additionally, the call avoids some table joins that can slow down the queries
in some cases.
As a result, the return value will no longer include the ``is_update`` field
when querying by ``rpmID``.
**Disable notifications by default in [un]tagBuildBypass calls**
@ -165,7 +169,7 @@ Tools that wish to generate email notifications will need to explicitly pass ``n
This option is a GLOB match pattern for the name of tag. You can now directly
call ``session.listTags(pattern='prefix-*-postfix')`` for example, to filter the
result list on server side. ``list-tags`` command tries its best to call it with
result list on server side. The ``list-tags`` command tries its best to call it with
``pattern`` as well.
@ -255,10 +259,16 @@ going to use the specfile with the SCM repo's name in root or ``SPECS`` dir.
| PR: https://pagure.io/koji/pull-request/2123
The ``preSCMCheckout`` and ``postSCMCheckout`` callbacks for kojid now include
a ``buildroot`` field that provides access to the internal ``BuildRoot``
object, when such an object is available.
This change impacts ``BuildMavenTask``, ``WrapperRPMTask``, ``ImageTask`` and
``BuildSRPMfromRPMTask``. Any plugins that use this should be aware that using
this could make them more fragile across releases. This feature does not come
with a promise avoid changing the behavior of the ``BuildRoot`` class.
``BuildSRPMfromRPMTask``.
The current exceptions are ``OzImageTask`` and ``BuildIndirectionImageTask``,
which do not use this type of buildroot.
Any plugins that use this field should be aware that the behavior of this class
may change across releases.
Web UI Changes
@ -302,6 +312,7 @@ Finally, python2 support for hub and web have been dropped in this release.
**Drop krbV support**
| PR: https://pagure.io/koji/pull-request/2244
| PR: https://pagure.io/koji/pull-request/2151
``krbV`` support has been finally removed from this release. For more information, please refer to
:ref:`migration_krbv`.
@ -326,26 +337,24 @@ benefit us for debugging purpose, as we are often using multicall more and more.
| PR: https://pagure.io/koji/pull-request/2308
| PR: https://pagure.io/koji/pull-request/2309
`PR#1417 <https://pagure.io/koji/pull-request/1417>`_ uses ``len()`` to check
the result of ``get_notification_recipients()``, but it could be ``None`` then
will cause a ``TypeError``. Now we fix this issue by both fixing the condition
in ``build_notification()`` and returning ``[]`` in
``get_notification_recipients()``.
This change fixes an inconsistency in the function where it would return
``None`` instead of an empty list as expected.
**Allow packagelist changes with 'tag' permission by the default policy**
| PR: https://pagure.io/koji/pull-request/2275
'tag' permission has been introduced for tag config management. It makes much
sense to let the users with `tag` permission be able to change packagelist as
well.
The ``tag`` permission was introduced in version 1.18 as part of an effort to
make admin permissions more granular.
This permission now grants access to make package list changes for tags
via the default ``package_list`` policy.
**Improve race condition for getNextRelease call and images**
| PR: https://pagure.io/koji/pull-request/2263
It was possible to meet the race condition in the old logic of image building.
We are now calling ``get_next_release()`` in ``initImageBuild`` call if there is
We are now calling ``get_next_release()`` in the ``initImageBuild`` call if there is
ino release passed in, rather than calling ``getNextRelease`` in the ImageBuild
task individually. This would notably reduce the possibility of the race
condition.
@ -362,7 +371,12 @@ but only keeping MD5 for RPM file processing.
| PR: https://pagure.io/koji/pull-request/2162
This option has been ``Off`` be default, see `mod_auth_gssapi doc
We have removed the ``GssapiSSLonly`` option from our example httpd
configuration.
It was previously shown in the example, set to ``Off``.
This is also the default in mod_auth_gssapi, but *it is not the recommended
setting*.
For more information, see `mod_auth_gssapi doc
<https://github.com/gssapi/mod_auth_gssapi#gssapisslonly>`_
**Remove "GssapiLocalName Off" option**
@ -370,7 +384,9 @@ This option has been ``Off`` be default, see `mod_auth_gssapi doc
| PR: https://pagure.io/koji/pull-request/2351
| PR: https://pagure.io/koji/pull-request/2358
_ditto_, and it is both for hub and web
We have also removed the ``GssapiLocalName`` option from our example httpd
configurations.
Similar to the above, our example setting was already the default.
**hub: Fix typo in ensure_volume_symlink**
@ -386,7 +402,7 @@ In particular, the ``buildtag`` policy test should work for such builds.
Note that some builds (e.g. content generator builds and other imported builds) do not
have associated tasks.
For more information on hub policies, see :doc:`defining_hub_policies`.
For more information on hub policies, see :doc:`../defining_hub_policies`.
**Archive's checksum_type should be always integer in DB**
@ -395,14 +411,11 @@ For more information on hub policies, see :doc:`defining_hub_policies`.
We fixed the problem in ``CG_Importer.match_file()`` and
``import_archive_internal()``.
**host.importImage doesn't honor volume**
**Honor volume policy in host.importImage**
| PR: https://pagure.io/koji/pull-request/2359
``host.importImage`` now directly uses the data of ``build_info`` rather than
fetching it from DB again. So, it won't miss the volume information anymore.
Notice that the signature has been changed: the argument ``build_id`` is changed to
``build_info``.
This fixes a bug where an underlying function as ignoring the volume policy result.
Plugins
@ -420,11 +433,17 @@ We now provide an easier way to find the owner of sidetags
**Give koji admins the permission to operate sidetags**
| PR: https://pagure.io/koji/pull-request/2322
Users with the ``admin`` permission can now manage sidetags even if they are
not their own.
**Fix is_sidetag_owner and is_sidetag policy tests**
| PR: https://pagure.io/koji/pull-request/2326
The admins should be able to manage sidetags even if they are not their own. This also
fix a bug that ``is_sidetag_owner``, ``is_sidetag`` used in policy check and many
other places do not return result.
These policy tests would previously always return a null result.
Now they return the correct one.
Utilities Changes
@ -433,12 +452,6 @@ Utilities Changes
Garbage Collector
.................
**Support of GSSAPI auth requests-kerberos**
| PR: https://pagure.io/koji/pull-request/2151
Meanwhile, the ``krb_login`` auth with ``krbV`` has been dropped.
**Systemd units for koji-gc**
| PR: https://pagure.io/koji/pull-request/2199
@ -457,9 +470,8 @@ New options ``cc_addr``, ``bcc_addr`` in config file, or CLI options
| PR: https://pagure.io/koji/pull-request/2253
The previous default value of ``smtp_host`` is ``None``. It will It will cause
``smtplib.SMTP().connect()`` to fail. Setting the default vault to 'localhost'
fixes this issue accordingly.
The previous the default value was ``None``, which would cause failures
if notifications were enabled.
Kojira
......
@ -476,9 +488,7 @@ this file in each cycle. For more information, please refer to
| PR: https://pagure.io/koji/pull-request/2154
``first_seen`` is measured from start of the kojira process. It doesn't make
much sense for short-lived kojira to determine the age of repo. Trying the best
to replace it with mtime of repo directory would be a more accurate approach.
Kojira should now do a better job of determining the age of a repo at startup.
**Fix logic detecting directories for pruneLocalRepos**
@ -507,8 +517,8 @@ Thus, we won't provide kojira service on <=EL6 platform.
| PR: https://pagure.io/koji/pull-request/2340
| PR: https://pagure.io/koji/pull-request/2397
Kojira are now able to delete repos in a separated thread. ``delete_batch_size``
is useless now.
Kojira are now able to delete repos in a separated thread.
The old ``delete_batch_size`` option is no longer used and has been removed.
koji-sidetag-cleanup
....................