Fix #4323: enter class type parameters in GADT bounds

[liufengyun]

Fix #4323: enter class type parameters in GADT bounds

[liufengyun]

test performance please

[dottybot]

performance test scheduled: 1 job(s) in queue, 0 running.

[dottybot]

Performance test finished successfully:

Visit http://dotty-bench.epfl.ch/4351/ to see the changes.

Benchmarks is based on merging with master (8feb596)

[allanrenucci]

Add a neg test where this.T is not other.T:

def join[A](other: Expr[A]): Expr[T] = ...

[smarter]

It'd be good to figure out why a test had to be moved to pos-deep-subtype, this often masks real bugs.

[Blaisorblade]

In fact the standard library had to be moved to -Yno-deep-subtypes, which seems even more concerning? We already know GADT bounds can create subtyping loops (#4069), so I'm not sure how robust this can possibly be.

[smarter]

If this is the same kind of loop as #4069 then we should probably try to fix the underlying issue first (I can give it a go).

[liufengyun]

For the test case tests/pos-deep-subtype/t5399.scala, I see following logs:

assertion failure for _ <:< c.T, frozen = true
assertion failure for _ <:< _, frozen = true
assertion failure for _ <:< c.T, frozen = true
assertion failure for _ <:< _, frozen = true
assertion failure for _ <:< c.T, frozen = true
assertion failure for _ <:< _, frozen = true
assertion failure for _ <:< c.T, frozen = true
assertion failure for _ <:< _, frozen = true
assertion failure for _ <:< c.T, frozen = true
assertion failure for _ <:< _, frozen = true
assertion failure for _ <:< c.T, frozen = true

@smarter I think it is the same subtyping loop, please go ahead.

[liufengyun]

Close for now, maybe @AleksanderBG can have a look at the GADTMap refactoring.