Implement the Validate RPC on built-in plugins

[edwbuck]

Update all plugins to have a Validate gRPC and centralize the existing validation logic such that it is reached from Validate and Configure gRPC calls.

[azdagron]

Something like this might help establish a pattern and reduce boilerplate: https://gist.github.com/azdagron/b55f1e70e1855cfa1aa73cb8f0027699

[edwbuck]

Something like this might help establish a pattern and reduce boilerplate: https://gist.github.com/azdagron/b55f1e70e1855cfa1aa73cb8f0027699

The pattern of coupling the configuration usability control and the note as an error or informational message under one API was adopted and applied to the plugins that have been reworked under this PR.

[amartinezfayo]

Hi @edwbuck, I spent some time looking at the aws_kms test failures.
I left some comments. In summary, I don't think that we need to make any change in the way that plugintest.Load is called in the setupTest function, so most of the changes can be avoided.

Here is a patch that essentially removes those changes and makes go test -race to pass.
aws_kms-test-fix.patch

[edwbuck]

Hi @edwbuck, I spent some time looking at the aws_kms test failures. I left some comments. In summary, I don't think that we need to make any change in the way that plugintest.Load is called in the setupTest function, so most of the changes can be avoided.

Here is a patch that essentially removes those changes and makes go test -race to pass. aws_kms-test-fix.patch

Thank you Agustin. I read over the patch, and in all other files I didn't notice that this combined two different testing patterns, such that it configured each plugin twice, creating the conditions that eventually triggered my race condition with my additional call to plugintest.CoreConfig(...).

Expect a push soon.

[edwbuck]

Ready for review two weeks ago. Please prioritize so I can stop keeping this up-to-date with the moving main branch.

[azdagron]

Thanks for putting in this work, @edwbuck. One overarching comment is that a large majority of the *_test.go changes could be reverted if we change the plugintest package to default to a valid CoreConfig (e.g. one with "example.org" provided as the trust domain). Tests that use plugintest.WithCoreConfig would still be able to override the default.