Fixing Trac #9046 (Panic in GHCi when using :print).
Hello, I am trying to fix the bug #9046 (Panic in GHCi when using :print). I've discovered that It was introduced by this SPJ's commit (25 Nov 2013): https://ghc.haskell.org/trac/ghc/changeset/a8ac471d435214dbdc1fa70f938c63128... and especially by this change: -type QuantifiedType = ([TyVar], Type) -- Make the free type variables explicit +type QuantifiedType = ([TyVar], Type) + -- Make the free type variables explicit + -- The returned Type should have no top-level foralls (I believe) quantifyType :: Type -> QuantifiedType --- Generalize the type: find all free tyvars and wrap in the appropiate ForAll. -quantifyType ty = (varSetElems (tyVarsOfType ty), ty) +-- Generalize the type: find all free and forall'd tyvars +-- and return them, together with the type inside, which +-- should not be a forall type. +-- +-- Thus (quantifyType (forall a. a->[b])) +-- returns ([a,b], a -> [b]) + +quantifyType ty = (varSetElems (tyVarsOfType rho), rho) + where + (_tvs, rho) = tcSplitForAllTys ty While this change looks reasonable to me it breaks exploiting thunks created by `:print` ghci's command from polymorphic values. Before this change we could have: $ ghci -fprint-explicit-foralls GHCi, version 7.6.3.20130421: http://www.haskell.org/ghc/ :? for help
:print length length = (_t1::forall a. [a] -> Int) _t1 [1,2,3] 3
And now we have: $ ghci -fprint-explicit-foralls GHCi, version 7.8.2: http://www.haskell.org/ghc/ :? for help
:print length length = (_t1::[a] -> Int) _t1 [1,2,3] ghc: panic! (the 'impossible' happened) (GHC version 7.8.2 for x86_64-unknown-linux): tcTyVarDetails a{tv arw} [tv]
This particular panic is caused by unification (compiler/typecheck/TcUnify.lhs line 1042). In case of `read` instead of `length` we get panic even earlier in simplification (compiler/typecheck/TcSimplify line 85). The concrete reason for the panic is wrong constructor used for tyvar (TyVar instead of TcTyVar). I see several ways to fix this: 1) rollback to forall-types in `:print` (requires change in cvObtainTerm, RtClosureInspect.hs); 2) skolemise free tyvar to RuntimeUnk (it becomes TcTyVar then): a) in cvObtainTerm; b) somewhere before typechecking. I can't decide the correct way to fix this bug or maybe there are other alternatives. Anyway I am ready to work on this though I need some advice. With best regards, Vitaly Bragilevsky
I'm not sure of the correct way to fix this either - Simon, what do you think? This is a regression in 7.8 relative to 7.6 which unfortunately wasn't caught by the test suite. The commit that broke this fixed something else, so it's not a simple matter of reverting it. Vitaly, one concrete thing you can do right now is add a test case and mark it broken by #9046. Cheers, Simon On 22/06/2014 15:47, Vitaly Bragilevsky wrote:
Hello,
I am trying to fix the bug #9046 (Panic in GHCi when using :print). I've discovered that It was introduced by this SPJ's commit (25 Nov 2013): https://ghc.haskell.org/trac/ghc/changeset/a8ac471d435214dbdc1fa70f938c63128... and especially by this change:
-type QuantifiedType = ([TyVar], Type) -- Make the free type variables explicit +type QuantifiedType = ([TyVar], Type) + -- Make the free type variables explicit + -- The returned Type should have no top-level foralls (I believe)
quantifyType :: Type -> QuantifiedType --- Generalize the type: find all free tyvars and wrap in the appropiate ForAll. -quantifyType ty = (varSetElems (tyVarsOfType ty), ty) +-- Generalize the type: find all free and forall'd tyvars +-- and return them, together with the type inside, which +-- should not be a forall type. +-- +-- Thus (quantifyType (forall a. a->[b])) +-- returns ([a,b], a -> [b]) + +quantifyType ty = (varSetElems (tyVarsOfType rho), rho) + where + (_tvs, rho) = tcSplitForAllTys ty
While this change looks reasonable to me it breaks exploiting thunks created by `:print` ghci's command from polymorphic values. Before this change we could have:
$ ghci -fprint-explicit-foralls GHCi, version 7.6.3.20130421: http://www.haskell.org/ghc/ :? for help
:print length length = (_t1::forall a. [a] -> Int) _t1 [1,2,3] 3
And now we have: $ ghci -fprint-explicit-foralls GHCi, version 7.8.2: http://www.haskell.org/ghc/ :? for help
:print length length = (_t1::[a] -> Int) _t1 [1,2,3] ghc: panic! (the 'impossible' happened) (GHC version 7.8.2 for x86_64-unknown-linux): tcTyVarDetails a{tv arw} [tv]
This particular panic is caused by unification (compiler/typecheck/TcUnify.lhs line 1042). In case of `read` instead of `length` we get panic even earlier in simplification (compiler/typecheck/TcSimplify line 85).
The concrete reason for the panic is wrong constructor used for tyvar (TyVar instead of TcTyVar).
I see several ways to fix this: 1) rollback to forall-types in `:print` (requires change in cvObtainTerm, RtClosureInspect.hs); 2) skolemise free tyvar to RuntimeUnk (it becomes TcTyVar then): a) in cvObtainTerm; b) somewhere before typechecking.
I can't decide the correct way to fix this bug or maybe there are other alternatives. Anyway I am ready to work on this though I need some advice.
With best regards, Vitaly Bragilevsky _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://www.haskell.org/mailman/listinfo/ghc-devs
participants (2)
-
Simon Marlow -
Vitaly Bragilevsky