[GHC] #13775: Type family expansion is too lazy, allows accepting of ill-typed terms

[GHC]

#13775: Type family expansion is too lazy, allows accepting of ill-typed terms
-------------------------------------+-------------------------------------
        Reporter:  fizruk            |                Owner:  diatchki
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler (Type    |              Version:  8.0.2
  checker)                           |             Keywords:
      Resolution:                    |  CustomTypeErrors
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC accepts       |  Unknown/Multiple
  invalid program                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by simonpj):

 * owner:  (none) => diatchki


Comment:

 This is really a question for Iavor, who implemented user-specified type
 errors.  The [http://downloads.haskell.org/~ghc/master/users-
 guide/glasgow_exts.html#custom-compile-time-errors user manual section on
 custom errors] does not explain the behaviour of the feature (except with
 a single example), and need some serious love.

 What is happening is this:

 * GHC only reports custom type errors if
   * We have an ''unsolved'' constraint involving `TypeError`
   * We have a declared or inferred type involving `TypeError`

 * We get a constraint `Show (Proxy (TypeError "..."))`.

 * But the instance for `Proxy` is
 {{{
 instance Show (Proxy s) where ...
 }}}
   so the `TypeError...` is discarded.

 * All constraints are solved, so no error is reported.

 Iavor: you may want to consider being more aggressive?  Or at least
 documenting the expected behaviour better.  Thanks!

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13775#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler