Phadej ghc 8.4

[ranjitjhala]

This PR moves LH to GHC 8.4 (thanks to @phadej -- cf #1330)

However, it is also a significant refactor of Bare, specifically, deleting BareM and all the inscrutable global dependencies that it wrought.

Not ready to merge yet.

• asserted-sigs
• unqualified imports
• assumed-sigs
• alias
• importing
• measures
• importing
• invariants
• importing
• holes
• reflect / PLE
• importing
• datacon
• importing
• class/instance measures (len)
• absref
• class-refinements
• lifted-bounds
• termination
• auto-size
• source-spans
• error-messages

Current Score (9/12/2018)

Only the following tests fail:

Fixpoint-apply-unknown-sort

• tests/todo-rebare/T1089b.hs [RJ]

Mystery

• tests/todo-rebare/DataKinds.hs [NV]

CLASS

• tests/todo-rebare/NatClass.hs [NV]
• tests/todo-rebare/VerifiedMonoid_NEG.hs [NV]
• tests/todo-rebare/VerifiedMonoid_POS.hs [NV]
• tests/classes/neg/Inst01.hs [NV]

[ranjitjhala]

Current score no-bench and no-error

Fri Aug 17 02:30:45 PDT 2018

rjhala@goto ~/r/s/l/liquidhaskell (phadej-ghc-8.4)> ack "OK" nohup.out  | wc
    862    2588   40670
rjhala@goto ~/r/s/l/liquidhaskell (phadej-ghc-8.4)> ack "FAIL" nohup.out  | wc
    177     540    8719

[ranjitjhala]

@nikivazou this is very close to merging.

I am going to remove the billion tracepp

scattered all over and then merge.

Better try it against your benchmarks soon :-)

[phadej]

@ranjitjhala FWIW, the dependencies are slowly but steadily catching up with GHC-8.6.1, so I might make another PR(s) after this is merged. (I'd probably wait for GHC-8.6.2 to come out first)

[ranjitjhala]

@phadej that would be great! Since stuff is relatively fresh in my mind, it should be possible to make that upgrade soon too!

[ranjitjhala]

Here goes!