[MISC] Update data-platform-libs: data_interfaces to 34 #220
Conversation
lucasgameiroborges
changed the title
| @patch_network_get(private_address="1.1.1.1") | ||
| @patch("charm.JujuVersion.has_secrets", new_callable=PropertyMock, return_value=True) | ||
| def test_migration_from_single_secret(self, scope, is_leader, _): | ||
| @pytest.mark.usefixtures("with_juju_secrets") |
There was a problem hiding this comment.
Apparently, juju secrets are not enabled in this test with the old patch, which is why I had to use the new fixture here (and also on canonical/pgbouncer-k8s-operator#277). Me and @shayancanonical investigated this issue, trying to figure out why the behavior is different, to no avail. If someone has an explanation to why that is, please let me know.
There was a problem hiding this comment.
It's been an issue I'm lobbying for from the start.
I stronlgy believe that ALL unittests must be run both with and without secrets.
I think on mysql-router we have a way to run them as such, in the same pipeline (allowing for coverage results collection -- which is normally the contra-argument why not to do it in the "easy way" in separate pieplines).
I can only ask for this once again... Guys please run ALL your unittests both with and without secrets.
It's virtually impossible to follow, which of your worklfows may or may not use them. Since peer data is used pretty much everywhere.
The safe way (and reaslitic feedback for coverage) is to run all unittests under both "secret-ful" and "secret-less".
lucasgameiroborges
requested review from
dragomirp,
marceloneppel and
taurus-forever
|
Duplicate of #207 |
|
Superseded by #207 |
No description provided.