Although I don't know anything about how the reporter is injected, that was another contributor who worked on that but I do remember it changing a few releases back when the nice coloring was added.
Hmm...no idea then. Just for troubleshooting purposes you could try loading an old pre-new reporter version, say 0.3.0 to see if that still has the issue. But it sounds related to the new reporter and I have no familiarity with that area but I may be able to help in a few days.
Either way, if it hurts I say it's worth opening an issue.
I’m going to drill into ensime first and if don’t make any progress I’ll come back to clippy
@crispywalrus we had a bit similar issue with SBT, as SBT was assuming a specific implementation of the reporter (which we have a custom implementation of). So maybe here the presentation compiler does the same - although I'm not using Ensime, so hard to say. If you won't find anything, please file a bug report :)
I'm using ensime SNAPSHOT so maybe it's something new :)
Defenitely something we need to look at. We managed to work around this in SBT with some dirty tricks (and it still happens but very rarely)
not nice when some code does a casting of a Reporter to some concrete type it expects it to be
I’ve moved the compiler version and clippy version around (couldn’t try the 0.3.0 suggestion as the code is build using 2.12 and used dependencies only available in 2.12 so …) it always comes down to https://github.com/scala/scala/blob/v2.12.1/src/compiler/scala/tools/nsc/Global.scala#L1519 or similar in the compiler during the presentation compiler run.
should InjectReporter be a GlobalPhase or is it somehow ending up in the list improperly?
higher up in the calls stack
interactive.Global there’s comments suggesting that sbt does inject phases into compilation and there’s some need for the compiler to be careful about this
In the end I can compile with clippy enabled but not use the presentation compiler … which breaks ensime
Our phases are quite simple
