refactor: replace ops-tracing vendored libs by Charmlibs versions

[dwilding]

Fully agent generated. I haven't reviewed yet.

[james-garner-canonical]

IIRC Ops needs to register its copy of the tracing library object under a different name to avoid clashing with the objects registered by charms using the same lib. I don't recall where we made this edit in the vendored lib or where we could make this adjustment to the PyPI lib (monkeypatch? subclass?). But I guess an agent should be able to take a crack at it.

Presumably this is why the observability tests fail.

[tonyandrewmeyer]

IIRC Ops needs to register its copy of the tracing library object under a different name to avoid clashing with the objects registered by charms using the same lib. I don't recall where we made this edit in the vendored lib or where we could make this adjustment to the PyPI lib (monkeypatch? subclass?). But I guess an agent should be able to take a crack at it.

Presumably this is why the observability tests fail.

I misremembered which library it was, it's certificate transfer, the "internal" prefix. And it is definitely the cause for the observability test failure. I would also assume subclassing or similar can work around it.

[dwilding]

That was it, thanks! My agent added subclasses in f0b625a. (Again, not yet reviewed by me)

[dwilding]

The Data Charm Tests failures are because the tracing lib was released too recently for Poetry (in downstream charm repos) to resolve. Should pass tomorrow.

[dwilding]

Hmm, the Data Charm Tests failures aren't as simple as I thought. postgresql-operator and postgresql-k8s-operator both require Pydantic 1, but the Charmlibs version of tracing drops support for Pydantic 1.

[tonyandrewmeyer]

Hmm, the Data Charm Tests failures aren't as simple as I thought. postgresql-operator and postgresql-k8s-operator both require Pydantic 1, but the Charmlibs version of tracing drops support for Pydantic 1.

For what it's worth, I handle that in my alternative to this approach. Maybe that would be useful for your agent in solving this? But I'm not totally sure, since the two approaches are quite different.