David M. Lloyd

Correct; independent keys can be placed into the same group, but they're still independent.

An initial naive implementation of this results in many errors from missing vtable and itable entries; looks like there's more to it than just this. I might poke at it...

Yeah I agree we're definitely reducing the set significantly between ADD and ANALYZE, but not enough. My initial testing with `grepj` show 856 reachable classes after ANALYZE after the naive...

This also may imply that we need a way to acquire the receiver's type ID from a "static" context, particularly if the type ID becomes an implicit parameter for these...

Maybe there should be a setting on the Diagnostic class itself for the stack trace. This way we can show partial stacks for the compiler and full stacks for the...

> It's not everything, but the core exception classes suggest that the revised reachability isn't seeing `VmHelpers.raiseAbstractMethodError()`, `VmHelpers.raiseNullPointerException`, etc. as being reachable code. I think that means the code isn't...

It might be a good idea to also add methods on to `Vm` to support allocating buffers and wrapping arrays as direct (possibly read-only) buffers as well.

@jaikiran commented [[email protected]](https://issues.redhat.com/secure/[email protected]), I will have some spare time this month and if this isn't something that you have already started work on, I can come up with an initial...

@dmlloyd commented Great! If this is done in the next few weeks, I'll be able to include it in the 1.6.0.Final release. Thanks!

@dmlloyd commented I'm definitely in favor of this enhancement. And, all the wiring should be there for this to be possible. But I am not familiar enough with the meaning...