Imported From: https://issues.scala-lang.org/browse/SI-7041?orig=1
Reporter: @adriaanm
See #6991

@xeno-by said:
Could we come up with a better name please? To tell the truth, I didn't understand what ConstantAnnotation is supposed to mean until I skimmed through the pull request discussion.

@lrytz said:
ideas?

@xeno-by said:
Reading this actually clarified things a lot, and now I think the name's fine:

I'm not sure if we should follow java here. Scala has only two levels: The default (same as SOURCE in java) and persisted (StaticAnnotation, same as RUNTIME in java). All persisted annotations are visible in (Scala) reflection (and of course in the compiler). I think this works well and I don't see much reason for adding a third level.

The thing we are discussing here is completely orthogonal: it's is about enforcing constant arguments. It's just a tool we give to annotation writers. They should be able to use it with normal and with static annotations.

Java compatibility (i.e. writing java-compatible annotations), finally, is yet another orthogonal issue. To emit java annotations we'd need to enforce constants, and on top of that we might also want to support the three java persistence levels. The latter could be achieved with meta-annotations. But let's not confuse things here.

There's another question though. Why do we need a hardcode into the compiler to verify that something's a constant, when we have macros? Unfortunately unless we have type macros, it's out of the question though :(

@lrytz said (edited on Feb 28, 2013 7:17:52 AM UTC):
This is on hold, it won't happen for 2.10.x because of forward binary compatibility, and in 2.11 we might have better ways to solve it (type macros). See https://groups.google.com/forum/#!topic/scala-internals/ipyrn6c8c8o/discussion

@adriaanm said:
Un-assigning to foster work stealing, as announced in https://groups.google.com/forum/?fromgroups=#!topic/scala-internals/o8WG4plpNkw

@soc said:
Is there a chance to get at least the existing work (remove the warning) merged?

I'm tired of getting spammed with 5 lines of nonsense every time I compile the library.

@lrytz said:
In my opinion, no. We can make the warning more precise, but that gives us even more noise in the places where it's being "abused".

@soc said:
What's more precise? As is, the warning is flat-out wrong.

@soc said:
Could we please please please get rid of this “2 plus 2 is not 5” warning?
I'm sick having to read this non-sense twice every time I compile Scala.

@adriaanm said:
Unassigning and rescheduling to M6 as previous deadline was missed.

@adriaanm said (edited on Jan 6, 2015 11:46:02 PM UTC):
We should really clean this up in 2.12. I find it quite bizarre that an annotation must be both java-defined and a subclass of ClassFileAnnotation (a Scala marker trait!). This conundrum was introduced 7 years ago, so it may be time to reconsider. (scala/scala@0b00cb9fc3)

@soc said:
Yep, I'm almost there (famous last words...)! I'm currently stuck on https://groups.google.com/d/msg/scala-internals/mH20kYunolM/93D41AjqdmAJ ... I either don't get it to work (condition doesn't trigger) or I run into an endless recursion. I need to look at it again to figure out what I'm doing wrong.

scala/scala#6143