[GHC] #12723: Family instance modules are not fingerprinted in ABI

[GHC]

#12723: Family instance modules are not fingerprinted in ABI
-------------------------------------+-------------------------------------
        Reporter:  ezyang            |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Driver            |              Version:  8.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Description changed by ezyang:

@@ -11,1 +11,1 @@
- module B where
+ module B (module A) where

New description:

 This leads to the following delightful, five module bug:

 {{{
 -- A.hs
 {-# LANGUAGE TypeFamilies #-}
 module A where
 type family F a
 type instance F Int = Bool

 -- B.hs
 module B (module A) where
 import A

 -- C.hs
 module C where
 import B

 -- D.hs
 module D where
 import C

 -- E.hs
 module E where
 import D
 import B
 }}}

 Build these modules. Then rename A to A2 (fixing B's import), and build
 them again. You'll get:

 {{{
 ezyang at sabre:~/Dev/labs/T3871$ ghc --make E.hs
 [1 of 5] Compiling A2               ( A2.hs, A2.o )
 [2 of 5] Compiling B                ( B.hs, B.o )
 [3 of 5] Compiling C                ( C.hs, C.o ) [B changed]
 [5 of 5] Compiling E                ( E.hs, E.o ) [B changed]
 attempting to use module ‘A’ (./A.hs) which is not loaded
 }}}

 The problem is clear: D was not recompiled, but it needs to be, because
 when the module gets renamed, we need to update its list of family
 instance modules to rename A to A2. When we don't do this, the subsequent
 family instance check chokes because it tries to load A. And why did D
 decide not to get recompiled? Because the ABI hash of C did not change.
 And that's WRONG.

 (Also, family instances are really awful, you really do have to rebuild
 everything when you change them. UGH.)

 This occasionally affected the Cabal project, see
 https://github.com/haskell/cabal/issues/3871

--

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12723#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler