Is it possible to have deprecated aliases?

I really believe we should completely remove it. I'm not convinced of any of the arguments I've heard so far...

Is it possible to have deprecated aliases?

It's been discussed, but we never implemented deprecating aliases, unfortunately.

I really believe we should completely remove it. I'm not convinced of any of the arguments I've heard so far...

It's extremely useful for unit tests that instantiate the compiler, in scala/scala there are lots of such uses. It would be cumbersome to manually assemble a symbol table classpath, we'd have to get information from the build tool to the test runtime.

Examples

• https://github.com/scala/scala/blob/v2.13.17/src/testkit/scala/tools/testkit/BytecodeTesting.scala#L168
• https://github.com/scala/scala/blob/v2.13.17/test/junit/scala/tools/nsc/symtab/SymbolTableForUnitTesting.scala#L77

It would be cumbersome

It would be cumbersome but it would be the correct usage of the compiler. We do that in Scala3. Vulpix doesn't use the underlying classpath and reconstruct them manually. See:

So if Scala 2 (years ago) did it for simplicity or whatever reason, I don't see it stand as a solid argument (Scala 3 took the time to have a correct testing suite).

It's a simple and reliable way to get a symbol classpath that's in sync with the library / compiler that we want to test, and we did it for 20 years without issues, one could see that as a solid argument in favor 😉

It's a simple and reliable way to get a symbol classpath that's in sync with the library / compiler that we want to test, and we did it for 20 years without issues, one could see that as a solid argument in favor 😉

I'm not saying it doesn't work. It does work to some extent. I'm saying it should have not existed in the first place and the underlying runtime is different from the compiler space. What if we are to compile the compiler with Scala.js and run it on the browser for scastie for example? How would this specific flag work and how it would be implemented when you don't have an underlying JVM?

...that aside, any clue about the scaladoc CI failure? 🤔

I understand your points, it seems we don't come to the same conclusion. I agree it should be, should always have been, a -Y flag.