Testsuite issues

[bgamari]

There are still a few test issues that could use some attention.

Redundant error messages

These tests (arguably) exhibit redundant error messages,

• [x] T6018ghcirnfail
• [x] tcfail058
• [x] tcfail078
• [x] tcfail063
• [x] tcfail113
• [x] tcfail161
• [x] tcfail212
• [x] T7609 RAE: Regrettably, this reports X a a where it should report aXa. But that requires a major, major overhaul to fix, I think. I don't spot anything redundant.

  Missing Expected/Actual clauses

• [x] ghci047 RAE: Ditto Defer02.
• [x] Defer02 RAE: When you look at the code, this makes sense, and I think is an improvement over HEAD. There is a function with a Int ~ Bool constraint. Expected/actual here is bogus.
• [x] T7453 RAE: The expected/actual is reported a bit further down. My branch finds this type error in a different spot than HEAD, but one isn't more right than the other. The expected/actual stuff is still reported; it just looks different.
• [x] FrozenErrorTests RAE: Like Defer02. I think this is an improvement.

  Other

• [x] drvfail009: This error message used to be more descriptive
• [x] T10267: Missing location information from TH splice
• [x] T7064: Missing type variable: GHC.Base.id :: forall . a_0 -> a_0
• [ ] T8583: Redundant reflective type equalities in context (e.g. Elem t ~ Elem t) RAE: This test does not exist. But note that HEAD also reports reflexive equalities when it has them.
• [x] T7778: Num Int => Num clearly nonsense, but I don't see the connection to impredicative polymorphism (although this may be more a reflection of my lacking type theory knowledge)
• [x] T9196: Error message used to be more descriptive (mention expected/actual kind of constraint)
• [x] T4007: Very suspicious: fold/build rule fires once instead of twice RAE: There is a story here, but it's a bit boring. But this is OK. In HEAD, you'll see that the rule fires only once also.
• [x] T7328: Missing location context "In the class declaration for Foo" RAE: This is a consequence of the design that errors are reported with 3 enclosing contexts, My error gives some more specific contexts, and so loses the outermost. I'm accepting the change.

  Compiler crashes

• [x] T7696

[goldfirere]

I'm closing this ticket, as I'm sweeping through the rest of the bugs. Thanks @bgamari for this!