Fix Oracle SSL tests

[jeff303]

Define new test for Oracle SSL connectivity, in oracle_test.clj, similar to how things work in mysql_test.clj

Add new test util macro, with-env-keys-renamed-by, to support running tests with environ keys temporarily renamed

Using new test macro from both MySQL and Oracle SSL connectivity tests

Removing now unneeded be-tests-oracle-ssl-ee CircleCI job

Removing now unneeded test-selector parameter for test-driver orbv in CircleCI config.yml

[jeff303]

Builds on top of #15172.

[codecov]

Codecov Report

Merging #15260 (b5995d0) into master (54f6a5e) will decrease coverage by 0.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #15260      +/-   ##
==========================================
- Coverage   84.39%   84.37%   -0.02%     
==========================================
  Files         398      398              
  Lines       31056    31056              
  Branches     2240     2239       -1     
==========================================
- Hits        26209    26204       -5     
- Misses       2607     2613       +6     
+ Partials     2240     2239       -1     

Impacted Files	Coverage Δ
src/metabase/driver/sql_jdbc/connection.clj	75.67% <0.00%> (-6.76%)	⬇️
src/metabase/models/query.clj	85.41% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 54f6a5e...b5995d0. Read the comment docs.

[dpsutton]

couple things about your macro with-env-keys-renamed-by
It iterates through every key. I think you could probably just leverage clojure.set/rename-keys

(tu/with-env-keys-whatever {:mb-mysql-test :mb-mysql-ssl-test} 
    (is (= ....))

Or you could take an arbitrary function f and call it on the env and redef with that value.

(tu/with-env-keys-whatever #(set/rename-keys % {:mb-mysql-test :mb-mysql-ssl-test})
  (is (= ...)))

That would allow more manipulation (if desired) than just renaming keys. You could duplicate, swap, what have you. But the api to take in every key as a string and return a new string seems a bit off.

Also, perhaps assert that the new env is different than the incoming environment? I imagine if you get this wrong and it doesn't update anything and your test fails for some weird reason it might be confusing. Or perhaps even worse, you get the key name wrong and your test still passes.

EDIT:
I just realized that you are swapping the prefix out, which is why you did that. Maybe that is the way to go. I think there might be some benefit to the rename keys where its more clear which keys are getting renamed but way less sure now. Feel free to ignore if none of this swayed you.

[jeff303]

couple things about your macro with-env-keys-renamed-by
It iterates through every key. I think you could probably just leverage clojure.set/rename-keys

(tu/with-env-keys-whatever {:mb-mysql-test :mb-mysql-ssl-test} 
    (is (= ....))

Or you could take an arbitrary function f and call it on the env and redef with that value.

(tu/with-env-keys-whatever #(set/rename-keys % {:mb-mysql-test :mb-mysql-ssl-test})
  (is (= ...)))

That would allow more manipulation (if desired) than just renaming keys. You could duplicate, swap, what have you. But the api to take in every key as a string and return a new string seems a bit off.

Also, perhaps assert that the new env is different than the incoming environment? I imagine if you get this wrong and it doesn't update anything and your test fails for some weird reason it might be confusing. Or perhaps even worse, you get the key name wrong and your test still passes.

EDIT:
I just realized that you are swapping the prefix out, which is why you did that. Maybe that is the way to go. I think there might be some benefit to the rename keys where its more clear which keys are getting renamed but way less sure now. Feel free to ignore if none of this swayed you.

Yeah, the intent was to have a convention that a certain prefix is renamed, in bulk, to cover things like variants of the same env (like SSL, or not, possibly other auth mechanisms or tweaks in the future). That's why I wanted to avoid having to individually enumerate the keys. However, the "by function" semantic is definitely compelling, as it is more generic (which is on brand for a test util macro). Will have to give it a bit more thought.

[camsaul]

1. You should alias namespaces you require and use those aliases, so it's clear at a glance that you're not using things without requiring them and causing namespace loading race conditions
2. New test util functions should be "imported" by Potemkin in metabase.test, and people using them should use them via mt/. Don't use tu/ or other stuff in new tests -- we don't want to be back in the place where people had to use 6 different test util namespaces to get anything done
3. Your macro should include some testing context to make debugging failing tests a little easier. Maybe you accidentally renamed all the env vars; you shouldn't need to debug the test util to figure that out -- it should tell you

[camsaul]

Cool LGTM