iceoryx
iceoryx copied to clipboard
eclipse-iceoryx
Remove all `TEST_ID`s and the `check_test_ids.sh`
Brief feature description
The TEST_IDs are no longer required for requirements tracing and can therefore be removed. The CI check script can be removed as well.
This would clean up the tests a bit.
@elfenpiff I'm all for it but what would be the alternative for requirement tracing? Do you suggest this is something downstream needs to figure out themselves?
@gpalmer-latai we briefly talked about entry hurdles into the project and you mentioned you know tests IDs from other projects. Maybe you can add your two cents here
Hm... I don't recall what comment I made to this effect. I've actually never encountered the idea of specifying Test IDs before. I could certainly ask around but I personally don't have a good enough picture of how the current system fulfills requirement tracing.
One dumb question - why not just generate the ID based on the test name? The name is unique so you can just use uuid5 rather having users manually create a uuid4.
I think what I might have said is that I'm familiar with the issue tracking process (linking issue numbers in PRs) and the github checklist and such. But specifically the TEST_ID invocation is a new concept to me. I don't actually know what useful thing generating a UUID does for a test. I could on the other hand imagine someone assigning an ID to a test because there was a requirement in JAMA. Or maybe folks want to add something to JAMA or some other tool after the fact - but again in this case I would imagine the test file + test name suffices to uniquely identify the test.
Okay actually, I HAVE had a brief discussion about this once with a colleague, talking about some testing framework.
The question was whether there should be some ID generated for tests and IIRC my conclusion was "Why? The name is already unique?" and also there wasn't a concrete requirement for the ID at the time, so the subject was quickly dropped.
@gpalmer-latai The idea was to decouple the test name which could change, e.g. new naming convention, from the actual identification and use an ID which does not change. There might be better solutions to this problem though.