typecheck/TcRnTypes panic caused by hi-boot error

[Simon Peyton-Jones]

I think I've fixed the "Missing top_errs' bug on the 6.02 branch.
(it's ok in the HEAD too)

Turned out to be 
a) A bad .hi file would provoke a crash rather than a more civilized
error.
b) The GLUT system could persuade GHC to make a worker for a default
	method with type	... -> forall a. (# ..., .. #)

The type checker thinks that a for-all with an unboxed inside is
erroneous,
and hence triggers (a).  

I've fixed both (a) and (b) on the branch.  (a) doesn't arise on the
HEAD,
because interface files aren't kind-checked any more.  (Except by lint)

Simon

| -----Original Message-----
| From: Simon Marlow
| Sent: 22 October 2003 14:19
| To: Simon Peyton-Jones
| Subject: FW: typecheck/TcRnTypes panic caused by hi-boot error
| 
| Hi Simon
| 
| George sent this to me, but you're probably in a better position to
deal with it...
| 
| -----Original Message-----
| From: George Russell [mailto:ger at tzi.de]
| Sent: 29 September 2003 14:48
| To: Simon Marlow
| Subject: typecheck/TcRnTypes panic caused by hi-boot error
| 
| Hello Simon,
| 
| the following type declaration in a .hi-boot file caused ghc-6.0.1 to
say
| "typecheck/TcRnTypes.lhs:233: Missing field in record construction
TcRnTypes.top_errs"
| 
| writeToMMiSSObject ::
|     MMiSSObjectTypeType.MMiSSObjectType
|     -> ViewType.View
|     -> MMiSSPackageFolder.MMiSSPackage.Folder
|     -> Data.Maybe.Maybe EntityNames.EntitySearchName
|     -> ELEMENT
|     -> GHC.Base.Bool
|     -> GHC.IOBase.IO (Computation.WithError (
|        Link.Link MMiSSObjectType.MMiSSObject,
|        Data.Maybe.Maybe ELEMENT))
| 
| 
| 
| Removing the erroneous "." in "MMiSSPackage.Folder" fixed it.
| 
| I don't want to send you all accompanying .hi files and so on because
there are
| probably about 400 of them.  Is this enough information for you?
| 
| best wishes,
| 
| George
|