Refactor runtime testsuite

[KvanTTT]

There is a lot of duplicated code in runtime test classes that can be extracted to a common abstract class. This issue can be assigned this issue to me.

[ericvergnaud]

I wouldn't be too worried about this.
I actually did the work some time ago and then dropped it, because I suspect it could on the long run create problems. Each target has specific requirements, and having the flexibility of changing the tests behavior for a specific target without having to think about other targets is highly valuable, especially given the specific skills of target authors.

[KvanTTT]

I don't completely agree. Most part of logic is common, specific parts can be extracted to specific classes. Also, it seems like some logic is excess.

I suspect it could on the long run create problems

If all tests are green after refactoring then there are no problems. Why type of problems do you mean?

Each target has specific requirements, and having the flexibility of changing the tests behavior for a specific target without having to think about other targets is highly valuable, especially given the specific skills of target authors.

As I've already said, specific logic can be successfully extracted to concrete classes, it's not a problem. Flexibility remains at the same level. I think we (maintainers and active contributors) shouldn't rely on target authors because they may suddenly leave the project. Moreover, they also may not know about all details (for instance, about optimizations in my latest PR).

Also, I would say the new structure of the tests suite will make it easier to add new targets and reduce the number of errors due to duplication.

[parrt]

per discussion #3369 easiest/least intrusive change is to use text blocks and convert comments with values into actual strings; tweaking the test classes (descriptors).

[KvanTTT]

It's about code itself, not about test data. It will allow us to add tests for new targets more easily and make code more clear. There is a lot of duplication now.

[parrt]

Resolved with #3407

[KvanTTT]

Actually not resolved. The issue is about something else, about code base, not about test data. Also, merge request is about data transformation, not refactoring.

[parrt]

Ah I see. Well, I'm going to close as the test rig works and there's little advantage to refactoring at this point. If we dramatically expand that area then it would be time to Refactor I would say.

[KvanTTT]

If we dramatically expand that area then it would be time to Refactor I would say.

It will allow adding new targets (Rust, maybe Kotlin) much easier. And will simplify code support.

[parrt]

Yes, that is true that adding new targets would be easier, but one of my rules for software development is to program for today and not tomorrow ;) we may never add a new target for example.

[parrt]

Change today is not less expensive than a change later when it's actually needed