Terrible failure of type inference in visible type application

Consider

-- optIntArg :: (Maybe Int -> t2) -> (t2,t2)
optIntArg f = (f Nothing, f (Just True))

This is rejected (by HEAD)

T11379a.hs:5:30: error:
    * Couldn't match type `a' with `Bool'
      `a' is a rigid type variable bound by
        a type expected by the context:
          forall a. Maybe a
        at T11379a.hs:5:30
      Expected type: forall a. Maybe a
        Actual type: Maybe Bool
    * In the first argument of `f', namely `(Just True)'
      In the expression: f (Just True)
      In the expression: (f Nothing, f (Just True))

but if you put the tuple components the other way round, it works fine

optIntArg f = (f (Just True), f Nothing)

Adding the commented-out signature also makes it work fine.

I'm almost certain that this is caused by visible type application; perhaps Nothing gets delayed instantiation, and then f's type becomes forall a. Maybe a. Utterly bogus.

I suspect it'll be fixed by Richards ReturnTv work, but let's make sure it is. We can't release this!!

Trac metadata

Trac field	Value
Version	7.10.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	highest
Resolution	Unresolved
Component	Compiler (Type checker)
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

mentioned in issue #11379 (closed)

changed milestone to %8.0.1

changed weight to 10

added Tbug Trac import typechecker labels

assigned to @rae

added TypeApplications label

Yes. This is a ReturnTv problem. Fix is nearly done -- I'm just going through the testsuite today.

mentioned in commit 00cbbab3

Trac metadata

Trac field	Value
Test case	- → typecheck/should_compile/T11458

added 1 deleted label

Trac metadata

Trac field	Value
Related	- → #11397 (closed)

closed

This was merged in 278e1fa2.

Trac metadata

Trac field	Value
Resolution	Unresolved → ResolvedFixed

removed 1 deleted label

mentioned in issue #9498

added Phighest label