Testcase tcfail158 is broken

Test testsuite/tests/typecheck/should_fail/tcfail158.hs contains the following code:

{-# LANGUAGE ExplicitForAll #-}

-- This one actually crashed in 6.4.1
-- There's a kind error in the signature for bar, 
-- and we were recovering, and then crashing when we found
-- a scoped type variable not in scope

 data Val v sm = Val
 foo :: forall v sm. Val v sm
 foo = undefined
   where foo1 :: Val v sm
         foo1 = bar
 -- Correct type signature: bar :: forall v sm. Val v sm
 bar :: forall v. Val v
 bar = undefined foo

The expected test result in testsuite/tests/typecheck/should_fail/tcfail158.stderr is:

tcfail158.hs:1:1: error:
    The IO action ‘main’ is not defined in module ‘Main’

Issues:

1. With a very high probability, the intention of this test was not to test on the missing main function. The intention was to test the error message for the type signature of function bar .
2. The code for the definitions and functions start in column 2 and not in column 1.
3. With an added main function GHC Versions 8.0, 8.2, 8.4 and 8.6 report the following error:

tcfail158.hs:17:18: error:
    • Expecting one more argument to ‘Val v’
      Expected a type, but ‘Val v’ has kind ‘* -> *’
    • In the type signature:
        bar :: forall v. Val v

1. GHC HEAD version 8.7.20181112, however, reports:

tcfail158.hs:15:17: error:
    • Couldn't match expected type ‘Val v1 sm1’
                  with actual type ‘Val v0’
    • In the expression: bar
      In an equation for ‘foo1’: foo1 = bar
      In an equation for ‘foo’:
          foo
            = undefined
            where
                foo1 :: Val v sm
                foo1 = bar
    • Relevant bindings include
        foo1 :: Val v1 sm1 (bound at tcfail158.hs:15:10)

tcfail158.hs:17:18: error:
    • Expecting one more argument to ‘Val v’
      Expected a type, but ‘Val v’ has kind ‘* -> *’
    • In the type signature: bar :: forall v. Val v
               ^^^^^
tcfail158.hs:18:7: error:
    • Couldn't match kind ‘* -> *’ with ‘*’
      When matching types
        t0 :: *
        Val v :: * -> *
    • In the expression: undefined foo
      In an equation for ‘bar’: bar = undefined foo
    • Relevant bindings include
        bar :: Val v (bound at tcfail158.hs:18:1)

GHC HEAD reports errors in 3 different lines 15, 17 and 18. Fix the type signature in line 17 and GHC compiles successfully. The error messages produced by GHC HEAD are very confusing!

Trac metadata

Trac field	Value
Version	8.7
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Test Suite
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

mentioned in issue #12906 (closed)

changed milestone to %8.8.1

changed weight to 5

added Tbug Trac import testsuite labels

It's worth noting that tcfail158 will in fact start reporting these errors once/if the fix for #12906 (closed) (Phab:D5338) lands.

Trac metadata

Trac field	Value
Related	- → #12906 (closed)

D5338 skips test tcfail158. Hence CI should not be affected.

Here is a somewhat simplified version of the issue:

module Bug where

f, g :: Maybe
f = Nothing
g = undefined

In GHC 8.6.2, this simply errors with:

$ ~/Software/haskell/ghc-8.6.2/bin/ghc Bug.hs
[1 of 1] Compiling Bug              ( Bug.hs, Bug.o )

Bug.hs:3:9: error:
    • Expecting one more argument to ‘Maybe’
      Expected a type, but ‘Maybe’ has kind ‘* -> *’
    • In the type signature: f :: Maybe
  |
3 | f, g :: Maybe
  |         ^^^^^

But in GHC HEAD (as of commit df570d92), we get many more errors:

$ ghc/inplace/bin/ghc-stage2 Bug.hs
[1 of 1] Compiling Bug              ( Bug.hs, Bug.o )

Bug.hs:3:9: error:
    • Expecting one more argument to ‘Maybe’
      Expected a type, but ‘Maybe’ has kind ‘* -> *’
    • In the type signature: f :: Maybe
  |
3 | f, g :: Maybe
  |         ^^^^^

Bug.hs:4:5: error:
    • Couldn't match expected type ‘Maybe’ with actual type ‘Maybe a1’
    • In the expression: Nothing
      In an equation for ‘f’: f = Nothing
  |
4 | f = Nothing
  |     ^^^^^^^

Bug.hs:5:5: error:
    • Couldn't match kind ‘* -> *’ with ‘*’
      When matching types
        a0 :: *
        Maybe :: * -> *
    • In the expression: undefined
      In an equation for ‘g’: g = undefined
  |
5 | g = undefined
  |     ^^^^^^^^^

I'll try tracking down which commit introduced this regression when I get a chance.

This regression was introduced in commit c955a514 (Remove decideKindGeneralisationPlan).

changed milestone to %8.10.1

Bumping milestones of low-priority tickets.

mentioned in commit 7eb4693b

mentioned in merge request !988 (closed)

As of current HEAD (7105fb66) the test gives a correct error message (see git log -p tcfail158.stderr) and the program from #15899 (comment 164036) no longer shows spurious kind error messages. I created !988 (closed) to re-enable the test.

I guess that commit 5c1f268e fixed the issue, but I haven't checked.

mentioned in commit fa0565df

mentioned in commit a0eae25d

mentioned in commit c5ee04fd

mentioned in commit d4738d00

mentioned in commit 40f37c43

mentioned in commit e3160ff9

mentioned in commit d30b550d

mentioned in commit feca4e1a

mentioned in commit 0b1fe737

mentioned in commit d31191c4

mentioned in commit 7ea2846f

mentioned in commit 162ada20

mentioned in commit 1f51aad6

closed

added Pnormal label