Jon Pretty

Yes, that's the sort of thing I was thinking of - sorry for not providing a more detailed description!

@odisseus It looks like your reproduction has a few absolute paths to JAR files in it, notably the Scala compiler JARs. I'm not sure the best way to make this...

I wonder if Scala considers the `match` to be a no-op rather than a cast because it's _statically_ true already. I'll try casting it to an object first to see...

One example is the work directory that's used to output classfiles to before they're packaged into a JAR. This directory is currently named after the project, module and schema (though...

This should wait for the new task system.

Sorry, I didn't fully explain the solution, as I was planning to experiment with a couple of possibilities... but hard links provide a convenient way to have multiple references to...

I thought about something like that, actually. It might help with invoking certain mutation actions, but wouldn't really solve handling output from Fury, though maybe it's still enough to make...

I guess this is because the *way* it writes files (create elsewhere, then move?) triggers a different inotify event from the one we're looking for.

The goal is actually to be better than Scalac's output. The output we get from Bloop (for Scala 2, at least) only gives us the start position of the error....

Or perhaps a cached indexed by a `Layer` instance.