December 2016 - ghc-tickets

[GHC] #12928: Too easy to trigger CUSK condition using TH
by GHC 09 Jul '18

09 Jul '18

#12928: Too easy to trigger CUSK condition using TH -------------------------------------+------------------------------------- Reporter: int-index | Owner: Type: feature | Status: new request | Priority: high | Milestone: Component: Compiler | Version: 8.0.1 (Type checker) | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: GHC rejects Unknown/Multiple | valid program Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- While trying to make `singletons` class promotion work with kind- polymorphic type variables, I've encountered a problem regarding CUSKs. Consider this class: {{{ $(promoteOnly [d| class Cls a where fff :: Proxy a -> () |]) }}} The generated TH (slightly simplified by hand) looks like this: {{{ type FffSym1 (t_alvm :: Proxy a1627472612) = Fff t_alvm data FffSym0 (l_alvn :: TyFun (Proxy a1627472612) ()) type instance Apply FffSym0 l_alvn = FffSym1 l_alvn class PCls a_alve where type Fff (arg_alvl :: Proxy a_alve) :: () }}} However, it fails to compile with the following error: {{{ cusk.hs:25:1: error: You have written a *complete user-suppled kind signature*, but the following variable is undetermined: k0 :: * Perhaps add a kind signature. Inferred kinds of user-written variables: a1627472612 :: k0 l_alvn :: TyFun (Proxy a1627472612) () }}} As suggested by the error message, we need to specify the kind in its entirety (write a proper CUSK). So to sidestep the issue, we can write {{{ data FffSym0 (l_alvn :: TyFun (Proxy (a1627472612 :: k)) ()) }}} and it'll work. However, this means that the TH code generator should perform some guesswork to modify kind annotations. It sounds like a minefield — too much can go wrong, so it'd be great to avoid doing so. Looking at the module `TcHsType` in the GHC sources, I see that the error is reported by the `kcHsTyVarBndrs` function: {{{ -- If there are any meta-tvs left, the user has -- lied about having a CUSK. Error. ; let (meta_tvs, good_tvs) = partition isMetaTyVar qkvs ; when (not (null meta_tvs)) $ report_non_cusk_tvs (qkvs ++ tc_tvs) }}} However, I don't see why we can't generalize over `meta_tvs` instead of reporting a failure. I asked on IRC #ghc and ezyang suggested that it should be sound. Below I attach a self-contained example that demonstrates the issue and has no dependencies except `base`. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12928> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 14

[GHC] #9123: Need for higher kinded roles
by GHC 08 Jul '18

08 Jul '18

#9123: Need for higher kinded roles ------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+------------------------------------- This [http://www.haskell.org/pipermail/ghc-devs/2014-May/004964.html thread] on ghc-devs identifies a real shortcoming of the new roles system. Here's a compact example, from the thread {{{ class C m where ret :: a -> m a bind :: m a -> (a -> m b) -> m b join :: m (m a) -> m a newtype T m a = MkT (m a) deriving( C ) }}} The `deriving(C)` is accepted without `join` in the class, but rejected when `join` is added. And the AMP proposal adds `join` to class `Monad`! In one way it is rightly rejected: it really would be unsound to derive an instance for `C (T K)` where `K`'s argument had nominal (but not representational) role. But we have no way to limit the type constructors at which `T` can be used. This deficiency is noted in the [http://research.microsoft.com/en- us/um/people/simonpj/papers/ext-f/ Safe Coercions paper], but this seems to be the first occasion on which it has bitten us badly. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9123> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 66

[GHC] #9672: Error message too long (full case statement printed)
by GHC 06 Jul '18

06 Jul '18

#9672: Error message too long (full case statement printed) -------------------------------------+------------------------------------- Reporter: andreas.abel | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler (Type | Version: 7.8.3 checker) | Operating System: Keywords: | Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: Other Difficulty: Easy (less than 1 | Test Case: hour) | Blocking: Blocked By: | Differential Revisions: Related Tickets: | -------------------------------------+------------------------------------- Imho, it does not make sense to print the other branches of a case when a type error is discovered in one branch. When you see the following example, you will agree with me. {{{ src/full/Agda/TypeChecking/Errors.hs:540:35: Couldn't match type ‘[Char]’ with ‘TCMT IO Doc’ Expected type: TCM Doc Actual type: [Char] In the expression: "Module cannot be imported since it has open interaction points" In a case alternative: SolvedButOpenHoles -> "Module cannot be imported since it has open interaction points" In a stmt of a 'do' block: }}} {{{#!hs case err of { InternalError s -> panic s NotImplemented s -> fwords $ "Not implemented: " ++ s NotSupported s -> fwords $ "Not supported: " ++ s CompilationError s -> sep [fwords "Compilation error:", text s] GenericError s -> fwords s GenericDocError d -> return d TerminationCheckFailed because -> fwords "Termination checking failed for the following functions:" $$ (nest 2 $ fsep $ punctuate comma $ map (pretty . dropTopLevelModule) $ concatMap termErrFunctions because) $$ fwords "Problematic calls:" $$ (nest 2 $ fmap (P.vcat . nub) $ mapM prettyTCM $ sortBy (compare `on` callInfoRange) $ concatMap termErrCalls because) PropMustBeSingleton -> fwords "Datatypes in Prop must have at most one constructor when proof irrelevance is enabled" DataMustEndInSort t -> fsep $ pwords "The type of a datatype must end in a sort." ++ [prettyTCM t] ++ pwords "isn't a sort." ShouldEndInApplicationOfTheDatatype t -> fsep $ pwords "The target of a constructor must be the datatype applied to its parameters," ++ [prettyTCM t] ++ pwords "isn't" ShouldBeAppliedToTheDatatypeParameters s t -> fsep $ pwords "The target of the constructor should be" ++ [prettyTCM s] ++ pwords "instead of" ++ [prettyTCM t] ShouldBeApplicationOf t q -> fsep $ pwords "The pattern constructs an element of" ++ [prettyTCM q] ++ pwords "which is not the right datatype" ShouldBeRecordType t -> fsep $ pwords "Expected non-abstract record type, found " ++ [prettyTCM t] ShouldBeRecordPattern p -> fsep $ pwords "Expected record pattern" NotAProjectionPattern p -> fsep $ pwords "Not a valid projection for a copattern: " ++ [prettyA p] DifferentArities -> fwords "The number of arguments in the defining equations differ" WrongHidingInLHS -> do { fwords "Unexpected implicit argument" } WrongHidingInLambda t -> do { fwords "Found an implicit lambda where an explicit lambda was expected" } WrongIrrelevanceInLambda t -> do { fwords "Found an irrelevant lambda where a relevant lambda was expected" } WrongNamedArgument a -> fsep $ pwords "Function does not accept argument " ++ [prettyTCM a] WrongHidingInApplication t -> do { fwords "Found an implicit application where an explicit application was expected" } HidingMismatch h h' -> fwords $ "Expected " ++ verbalize (Indefinite h') ++ " argument, but found " ++ verbalize (Indefinite h) ++ " argument" RelevanceMismatch r r' -> fwords $ "Expected " ++ verbalize (Indefinite r') ++ " argument, but found " ++ verbalize (Indefinite r) ++ " argument" ColorMismatch c c' -> fsep $ pwords "Expected argument color to be" ++ [prettyTCM c'] ++ pwords "but found color" ++ [prettyTCM c] NotInductive t -> fsep $ [prettyTCM t] ++ pwords "is not an inductive data type" UninstantiatedDotPattern e -> fsep $ pwords "Failed to infer the value of dotted pattern" IlltypedPattern p a -> fsep $ pwords "Type mismatch" IllformedProjectionPattern p -> fsep $ pwords "Ill-formed projection pattern " ++ [prettyA p] CannotEliminateWithPattern p a -> do { let ...; fsep $ pwords "Cannot eliminate type" ++ prettyTCM a : if isProj then pwords "with projection pattern" ++ ... else pwords "with pattern" ++ prettyA p : pwords "(did you supply too many arguments?)" } TooManyArgumentsInLHS a -> fsep $ pwords "Left hand side gives too many arguments to a function of type" ++ [prettyTCM a] WrongNumberOfConstructorArguments c expect given -> fsep $ pwords "The constructor" ++ [prettyTCM c] ++ pwords "expects" ++ [text (show expect)] ++ pwords "arguments (including hidden ones), but has been given" ++ [text (show given)] ++ pwords "(including hidden ones)" CantResolveOverloadedConstructorsTargetingSameDatatype d cs -> fsep $ pwords ("Can't resolve overloaded constructors targeting the same datatype (" ++ show (qnameToConcrete d) ++ "):") ++ map pretty cs DoesNotConstructAnElementOf c t -> fsep $ pwords "The constructor" ++ [prettyTCM c] ++ pwords "does not construct an element of" ++ [prettyTCM t] ConstructorPatternInWrongDatatype c d -> fsep $ [prettyTCM c] ++ pwords "is not a constructor of the datatype" ++ [prettyTCM d] IndicesNotConstructorApplications [i] -> fwords "The index" $$ nest 2 (prettyTCM i) $$ fsep (pwords "is not a constructor (or literal) applied to variables" ++ pwords "(note that parameters count as constructor arguments)") IndicesNotConstructorApplications is -> fwords "The indices" $$ nest 2 (vcat $ map prettyTCM is) $$ fsep (pwords "are not constructors (or literals) applied to variables" ++ pwords "(note that parameters count as constructor arguments)") IndexVariablesNotDistinct vs is -> fwords "The variables" $$ nest 2 (vcat $ map (\ v -> prettyTCM (I.Var v [])) vs) $$ fwords "in the indices" $$ nest 2 (vcat $ map prettyTCM is) $$ fwords "are not distinct (note that parameters count as constructor arguments)" IndicesFreeInParameters vs indices pars -> fwords "The variables" $$ nest 2 (vcat $ map (\ v -> prettyTCM (I.Var v [])) vs) $$ fwords "which are used (perhaps as constructor parameters) in the index expressions" $$ nest 2 (vcat $ map prettyTCM indices) $$ fwords "are free in the parameters" $$ nest 2 (vcat $ map prettyTCM pars) ShadowedModule x [] -> (throwImpossible (Impossible "src/full/Agda/TypeChecking/Errors.hs" 404)) ShadowedModule x ms@(m : _) -> fsep $ pwords "Duplicate definition of module" ++ [prettyTCM x <> text "."] ++ pwords "Previous definition of" ++ [help m] ++ pwords "module" ++ [prettyTCM x] ++ pwords "at" ++ [prettyTCM r] where help m = do { ... } r = case ... of { [] -> ... r : _ -> ... } defSiteOfLast [] = noRange defSiteOfLast ns = nameBindingSite (last ns) ModuleArityMismatch m EmptyTel args -> fsep $ pwords "The module" ++ [prettyTCM m] ++ pwords "is not parameterized, but is being applied to arguments" ModuleArityMismatch m tel@(ExtendTel _ _) args -> fsep $ pwords "The arguments to " ++ [prettyTCM m] ++ pwords "does not fit the telescope" ++ [prettyTCM tel] ShouldBeEmpty t [] -> fsep $ [prettyTCM t] ++ pwords "should be empty, but that's not obvious to me" ShouldBeEmpty t ps -> fsep ([prettyTCM t] ++ pwords "should be empty, but the following constructor patterns are valid:") $$ nest 2 (vcat $ map (showPat 0) ps) ShouldBeASort t -> fsep $ [prettyTCM t] ++ pwords "should be a sort, but it isn't" ShouldBePi t -> fsep $ [prettyTCM t] ++ pwords "should be a function type, but it isn't" NotAProperTerm -> fwords "Found a malformed term" SetOmegaNotValidType -> fwords "Set\969 is not a valid type" InvalidType v -> fsep $ [prettyTCM v] ++ pwords "is not a valid type" SplitOnIrrelevant p t -> fsep $ pwords "Cannot pattern match" ++ [prettyA p] ++ pwords "against irrelevant type" ++ [prettyTCM t] DefinitionIsIrrelevant x -> fsep $ text "Identifier" : prettyTCM x : pwords "is declared irrelevant, so it cannot be used here" VariableIsIrrelevant x -> fsep $ text "Variable" : prettyTCM x : pwords "is declared irrelevant, so it cannot be used here" UnequalBecauseOfUniverseConflict cmp s t -> fsep $ [prettyTCM s, notCmp cmp, ....] UnequalTerms cmp s t a -> do { (d1, d2, d) <- prettyInEqual s t; fsep $ [...] ++ pwords "of type" ++ [...] ++ [...] } UnequalTypes cmp a b -> prettyUnequal a (notCmp cmp) b UnequalColors a b -> error "TODO guilhem 4" HeterogeneousEquality u a v b -> fsep $ pwords "Refuse to solve heterogeneous constraint" ++ [prettyTCM u] ++ pwords ":" ++ [prettyTCM a] ++ pwords "=?=" ++ [prettyTCM v] ++ pwords ":" ++ [prettyTCM b] UnequalRelevance cmp a b -> fsep $ [prettyTCM a, notCmp cmp, ....] ++ pwords "because one is a relevant function type and the other is an irrelevant function type" UnequalHiding a b -> fsep $ [prettyTCM a, text "!=", ....] ++ pwords "because one is an implicit function type and the other is an explicit function type" UnequalSorts s1 s2 -> fsep $ [prettyTCM s1, text "!=", ....] NotLeqSort s1 s2 -> fsep $ pwords "The type of the constructor does not fit in the sort of the datatype, since" ++ [prettyTCM s1] ++ pwords "is not less or equal than" ++ [prettyTCM s2] TooFewFields r xs -> fsep $ pwords "Missing fields" ++ punctuate comma (map pretty xs) ++ pwords "in an element of the record" ++ [prettyTCM r] TooManyFields r xs -> fsep $ pwords "The record type" ++ [prettyTCM r] ++ pwords "does not have the fields" ++ punctuate comma (map pretty xs) DuplicateConstructors xs -> fsep $ pwords "Duplicate constructors" ++ punctuate comma (map pretty xs) ++ pwords "in datatype" DuplicateFields xs -> fsep $ pwords "Duplicate fields" ++ punctuate comma (map pretty xs) ++ pwords "in record" UnexpectedWithPatterns ps -> fsep $ pwords "Unexpected with patterns" ++ (punctuate (text " |") $ map prettyA ps) WithClausePatternMismatch p q -> fsep $ pwords "With clause pattern " ++ [prettyA p] ++ pwords " is not an instance of its parent pattern " ++ [prettyTCM q] MetaCannotDependOn m ps i -> fsep $ pwords "The metavariable" ++ [prettyTCM $ MetaV m []] ++ pwords "cannot depend on" ++ [pvar i] ++ pwords "because it" ++ deps where pvar = prettyTCM . var deps = case map pvar ps of { [] -> ... [x] -> ... xs -> ... } MetaOccursInItself m -> fsep $ pwords "Cannot construct infinite solution of metavariable" ++ [prettyTCM $ MetaV m []] BuiltinMustBeConstructor s e -> fsep $ [prettyA e] ++ pwords "must be a constructor in the binding to builtin" ++ [text s] NoSuchBuiltinName s -> fsep $ pwords "There is no built-in thing called" ++ [text s] DuplicateBuiltinBinding b x y -> fsep $ pwords "Duplicate binding for built-in thing" ++ [text b <> comma] ++ pwords "previous binding to" ++ [prettyTCM x] NoBindingForBuiltin x -> fsep $ pwords "No binding for builtin thing" ++ [text x <> comma] ++ pwords ("use {-# BUILTIN " ++ x ++ " name #-} to bind it to 'name'") NoSuchPrimitiveFunction x -> fsep $ pwords "There is no primitive function called" ++ [text x] BuiltinInParameterisedModule x -> fwords $ "The BUILTIN pragma cannot appear inside a bound context " ++ "(for instance, in a parameterised module or as a local declaration)" IllegalLetInTelescope tb -> fsep $ [pretty tb] ++ pwords " is not allowed in a telescope here." NoRHSRequiresAbsurdPattern ps -> fwords $ "The right-hand side can only be omitted if there " ++ "is an absurd pattern, () or {}, in the left-hand side." AbsurdPatternRequiresNoRHS ps -> fwords $ "The right-hand side must be omitted if there " ++ "is an absurd pattern, () or {}, in the left-hand side." LocalVsImportedModuleClash m -> fsep $ pwords "The module" ++ [text $ show m] ++ pwords "can refer to either a local module or an imported module" SolvedButOpenHoles -> "Module cannot be imported since it has open interaction points" UnsolvedMetas rs -> fsep (pwords "Unsolved metas at the following locations:") $$ nest 2 (vcat $ map prettyTCM rs) UnsolvedConstraints cs -> fsep (pwords "Failed to solve the following constraints:") $$ nest 2 (vcat $ map prettyConstraint cs) where prettyConstraint :: ProblemConstraint -> TCM Doc prettyConstraint c = f (prettyTCM c) where r = ... .... CyclicModuleDependency ms -> fsep (pwords "cyclic module dependency:") $$ nest 2 (vcat $ map pretty ms) FileNotFound x files -> fsep (pwords "Failed to find source of module" ++ [pretty x] ++ pwords "in any of the following locations:") $$ nest 2 (vcat $ map (text . filePath) files) OverlappingProjects f m1 m2 -> fsep (pwords "The file" ++ [text (filePath f)] ++ pwords "can be accessed via several project roots. Both" ++ [pretty m1] ++ pwords "and" ++ [pretty m2] ++ pwords "point to this file.") AmbiguousTopLevelModuleName x files -> fsep (pwords "Ambiguous module name. The module name" ++ [pretty x] ++ pwords "could refer to any of the following files:") $$ nest 2 (vcat $ map (text . filePath) files) ClashingFileNamesFor x files -> fsep (pwords "Multiple possible sources for module" ++ [text $ show x] ++ pwords "found:") $$ nest 2 (vcat $ map (text . filePath) files) ModuleDefinedInOtherFile mod file file' -> fsep $ pwords "You tried to load" ++ [text (filePath file)] ++ pwords "which defines the module" ++ [pretty mod <> text "."] ++ pwords "However, according to the include path this module should" ++ pwords "be defined in" ++ [text (filePath file') <> text "."] ModuleNameDoesntMatchFileName given files -> fsep (pwords "The name of the top level module does not match the file name. The module" ++ [pretty given] ++ pwords "should be defined in one of the following files:") $$ nest 2 (vcat $ map (text . filePath) files) BothWithAndRHS -> fsep $ pwords "Unexpected right hand side" NotInScope xs -> fsep (pwords "Not in scope:") $$ nest 2 (vcat $ map name xs) where name x = fsep [...] suggestion s | elem ':' s = parens $ text "did you forget space around the ':'?" | elem "->" two = parens $ text "did you forget space around the '->'?" | otherwise = empty where two = ... NoSuchModule x -> fsep $ pwords "No such module" ++ [pretty x] AmbiguousName x ys -> vcat [fsep $ pwords "Ambiguous name" ++ [...] ++ pwords "It could refer to any one of", nest 2 $ vcat $ map nameWithBinding ys, ....] AmbiguousModule x ys -> vcat [fsep $ pwords "Ambiguous module name" ++ [...] ++ pwords "It could refer to any one of", nest 2 $ vcat $ map help ys, ....] where help :: ModuleName -> TCM Doc help m = do { ... } UninstantiatedModule x -> fsep (pwords "Cannot access the contents of the parameterised module" ++ [pretty x <> text "."] ++ pwords "To do this the module first has to be instantiated. For instance:") $$ nest 2 (hsep [text "module", pretty x <> text "'", ....]) ClashingDefinition x y -> fsep $ pwords "Multiple definitions of" ++ [pretty x <> text "."] ++ pwords "Previous definition at" ++ [prettyTCM $ nameBindingSite $ qnameName y] ClashingModule m1 m2 -> fsep $ pwords "The modules" ++ [prettyTCM m1, text "and", ....] ++ pwords "clash." ClashingImport x y -> fsep $ pwords "Import clash between" ++ [pretty x, text "and", ....] ClashingModuleImport x y -> fsep $ pwords "Module import clash between" ++ [pretty x, text "and", ....] PatternShadowsConstructor x c -> fsep $ pwords "The pattern variable" ++ [prettyTCM x] ++ pwords "has the same name as the constructor" ++ [prettyTCM c] DuplicateImports m xs -> fsep $ pwords "Ambiguous imports from module" ++ [pretty m] ++ pwords "for" ++ punctuate comma (map pretty xs) ModuleDoesntExport m xs -> fsep $ pwords "The module" ++ [pretty m] ++ pwords "doesn't export the following:" ++ punctuate comma (map pretty xs) NotAModuleExpr e -> fsep $ pwords "The right-hand side of a module definition must have the form 'M e1 .. en'" ++ pwords "where M is a module name. The expression" ++ [pretty e, text "doesn't."] FieldOutsideRecord -> fsep $ pwords "Field appearing outside record declaration." InvalidPattern p -> fsep $ pretty p : pwords "is not a valid pattern" RepeatedVariablesInPattern xs -> fsep $ pwords "Repeated variables in left hand side:" ++ map pretty xs NotAnExpression e -> fsep $ [pretty e] ++ pwords "is not a valid expression." NotAValidLetBinding nd -> fwords $ "Not a valid let-declaration" NothingAppliedToHiddenArg e -> fsep $ [pretty e] ++ pwords "cannot appear by itself. It needs to be the argument to" ++ pwords "a function expecting an implicit argument." NothingAppliedToInstanceArg e -> fsep $ [pretty e] ++ pwords "cannot appear by itself. It needs to be the argument to" ++ pwords "a function expecting an instance argument." NoParseForApplication es -> fsep $ pwords "Could not parse the application" ++ [pretty $ C.RawApp noRange es] AmbiguousParseForApplication es es' -> fsep (pwords "Don't know how to parse" ++ [pretty_es <> (text ".")] ++ pwords "Could mean any one of:") $$ nest 2 (vcat $ map pretty' es') where pretty_es :: TCM Doc pretty_es = pretty $ C.RawApp noRange es pretty' :: C.Expr -> TCM Doc .... BadArgumentsToPatternSynonym x -> fsep $ pwords "Bad arguments to pattern synonym " ++ [prettyTCM x] TooFewArgumentsToPatternSynonym x -> fsep $ pwords "Too few arguments to pattern synonym " ++ [prettyTCM x] UnusedVariableInPatternSynonym -> fsep $ pwords "Unused variable in pattern synonym." NoParseForLHS IsLHS p -> fsep $ pwords "Could not parse the left-hand side" ++ [pretty p] NoParseForLHS IsPatSyn p -> fsep $ pwords "Could not parse the pattern synonym" ++ [pretty p] AmbiguousParseForLHS lhsOrPatSyn p ps -> fsep (pwords "Don't know how to parse" ++ [pretty_p <> text "."] ++ pwords "Could mean any one of:") $$ nest 2 (vcat $ map pretty' ps) where pretty_p :: TCM Doc pretty_p = pretty p pretty' :: C.Pattern -> TCM Doc .... IncompletePatternMatching v args -> fsep $ pwords "Incomplete pattern matching for" ++ [prettyTCM v <> text "."] ++ pwords "No match for" ++ map prettyTCM args UnreachableClauses f pss -> fsep $ pwords "Unreachable" ++ pwords (plural (length pss) "clause") where plural 1 thing = thing plural n thing = thing ++ "s" CoverageFailure f pss -> fsep (pwords "Incomplete pattern matching for" ++ [prettyTCM f <> text "."] ++ pwords "Missing cases:") $$ nest 2 (vcat $ map display pss) where display ps = do { ... } nicify f ps = do { ... } CoverageCantSplitOn c tel cIxs gIxs | length cIxs /= length gIxs -> (throwImpossible (Impossible "src/full/Agda/TypeChecking/Errors.hs" 750)) | otherwise -> addCtxTel tel $ vcat ([fsep $ pwords "I'm not sure if there should be a case for the constructor" ++ [...] ++ pwords "because I get stuck when trying to solve the following" ++ pwords "unification problems (inferred index \8799 expected index):"] ++ zipWith (\ c g -> nest 2 $ prettyTCM c <+> text "\8799" <+> prettyTCM g) cIxs gIxs) CoverageCantSplitIrrelevantType a -> fsep $ pwords "Cannot split on argument of irrelevant datatype" ++ [prettyTCM a] CoverageCantSplitType a -> fsep $ pwords "Cannot split on argument of non-datatype" ++ [prettyTCM a] SplitError e -> prettyTCM e WithoutKError a u v -> fsep $ pwords "Cannot eliminate reflexive equation" ++ [prettyTCM u] ++ pwords "=" ++ [prettyTCM v] ++ pwords "of type" ++ [prettyTCM a] ++ pwords "because K has been disabled." NotStrictlyPositive d ocs -> fsep $ pwords "The datatype" ++ [prettyTCM d] ++ pwords "is not strictly positive, because" ++ prettyOcc "it" ocs where prettyOcc _ [] = [] prettyOcc it (OccCon d c r : ocs) = concat [...] prettyOcc it (OccClause f n r : ocs) = concat [...] prettyR NonPositively = pwords "negatively" prettyR (ArgumentTo i q) = pwords "as the" ++ [...] ++ pwords "argument to" ++ [...] th 0 = text "first" th 1 = text "second" th 2 = text "third" th n = text (show $ n - 1) <> text "th" .... IFSNoCandidateInScope t -> fsep $ pwords "No variable of type" ++ [prettyTCM t] ++ pwords "was found in scope." SafeFlagPostulate e -> fsep $ pwords "Cannot postulate" ++ [pretty e] ++ pwords "with safe flag" SafeFlagPragma xs -> let plural | length xs == 1 = ... | otherwise = ... in fsep $ [fwords ("Cannot set OPTION pragma" ++ plural)] ++ map text xs ++ [fwords "with safe flag."] SafeFlagNoTerminationCheck -> fsep (pwords "Cannot use NO_TERMINATION_CHECK pragma with safe flag.") SafeFlagNonTerminating -> fsep (pwords "Cannot use NON_TERMINATING pragma with safe flag.") SafeFlagTerminating -> fsep (pwords "Cannot use TERMINATING pragma with safe flag.") SafeFlagPrimTrustMe -> fsep (pwords "Cannot use primTrustMe with safe flag") NeedOptionCopatterns -> fsep (pwords "Option --copatterns needed to enable destructor patterns") } }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9672> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #12073: Missing instance of MonadFix for Q
by GHC 06 Jul '18

06 Jul '18

#12073: Missing instance of MonadFix for Q -------------------------------------+------------------------------------- Reporter: jophish | Owner: Type: feature | Status: new request | Priority: normal | Milestone: Component: Template | Version: Haskell | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- `Q` is capable of being an instance of `MonadFix`. The definition of `mfix` is the same as `fixIO` except that the `MVar` handling is wrapped in `runIO`. {{{#!hs instance MonadFix Q where mfix k = do m <- runIO newEmptyMVar ans <- runIO (unsafeInterleaveIO (takeMVar m)) result <- k ans runIO (putMVar m result) pure result }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12073> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #11523: Infinite Loop when mixing UndecidableSuperClasses and the class/instance constraint synonym trick.
by GHC 05 Jul '18

05 Jul '18

#11523: Infinite Loop when mixing UndecidableSuperClasses and the class/instance constraint synonym trick. -------------------------------------+------------------------------------- Reporter: ekmett | Owner: Type: bug | Status: new Priority: highest | Milestone: 8.0.1 Component: Compiler | Version: 7.10.3 (Type checker) | Keywords: | Operating System: Unknown/Multiple UndecidableSuperClasses | Architecture: | Type of failure: GHC rejects Unknown/Multiple | valid program Test Case: | Blocked By: Blocking: | Related Tickets: #11480 Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- There seems to be a bad interaction between `UndecidableSuperClasses` and the common trick of using a cyclic definition of a class and instance to make an alias at the constraint level: {{{#!hs class (Foo f, Bar f) => Baz f instance (Foo f, Bar f) => Baz f }}} Unfortunately, there are circumstances in which you can't eliminate it, such as {{{#!hs class (p, q) => p & q instance (p, q) => p & q }}} There we can't partially apply (,) at the constraint level, but we can talk about `(&) :: Constraint -> Constraint -> Constraint` and `(&) (Eq a) :: Constraint -> Constraint`. This doesn't seem to happen on simpler examples like the above, but once I modify the categories example from #11480 to move the domain and codomain of a functor into class associated types, so that they don't infect every single subclass of functor, we run into a problem. The following stripped down version of the code seems to send the compiler into an infinite loop: {{{#!hs {-# language KindSignatures #-} {-# language PolyKinds #-} {-# language DataKinds #-} {-# language TypeFamilies #-} {-# language RankNTypes #-} {-# language NoImplicitPrelude #-} {-# language FlexibleContexts #-} {-# language MultiParamTypeClasses #-} {-# language GADTs #-} {-# language ConstraintKinds #-} {-# language FlexibleInstances #-} {-# language TypeOperators #-} {-# language ScopedTypeVariables #-} {-# language DefaultSignatures #-} {-# language FunctionalDependencies #-} {-# language UndecidableSuperClasses #-} {-# language UndecidableInstances #-} {-# language TypeInType #-} import GHC.Types (Constraint, Type) import qualified Prelude type Cat i = i -> i -> Type newtype Y (p :: i -> j -> Type) (a :: j) (b :: i) = Y { getY :: p b a } class Vacuous (a :: i) instance Vacuous a class (Functor p, Dom p ~ Op p, Cod p ~ Nat p (->)) => Category (p :: Cat i) where type Op p :: Cat i type Op p = Y p type Ob p :: i -> Constraint type Ob p = Vacuous class (Category (Dom f), Category (Cod f)) => Functor (f :: i -> j) where type Dom f :: Cat i type Cod f :: Cat j class (Functor f, Dom f ~ p, Cod f ~ q) => Fun (p :: Cat i) (q :: Cat j) (f :: i -> j) | f -> p q instance (Functor f, Dom f ~ p, Cod f ~ q) => Fun (p :: Cat i) (q :: Cat j) (f :: i -> j) data Nat (p :: Cat i) (q :: Cat j) (f :: i -> j) (g :: i -> j) instance (Category p, Category q) => Category (Nat p q) where type Ob (Nat p q) = Fun p q instance (Category p, Category q) => Functor (Nat p q) where type Dom (Nat p q) = Y (Nat p q) type Cod (Nat p q) = Nat (Nat p q) (->) instance (Category p, Category q) => Functor (Nat p q f) where type Dom (Nat p q f) = Nat p q type Cod (Nat p q f) = (->) instance Category (->) instance Functor ((->) e) where type Dom ((->) e) = (->) type Cod ((->) e) = (->) instance Functor (->) where type Dom (->) = Y (->) type Cod (->) = Nat (->) (->) instance (Category p, Op p ~ Y p) => Category (Y p) where type Op (Y p) = p instance (Category p, Op p ~ Y p) => Functor (Y p a) where type Dom (Y p a) = Y p type Cod (Y p a) = (->) instance (Category p, Op p ~ Y p) => Functor (Y p) where type Dom (Y p) = p type Cod (Y p) = Nat (Y p) (->) }}} Here I need the circular definition of `Fun` to talk about the fact that the objects in the category of natural transformations from a category p to a category q are functors with domain p and codomain q, so to give the definition of the class-associated `type Ob (Nat p q)` I need such a cyclic definition. I can't leave the domain and codomain of `Functor` in fundeps, otherwise if I go to define a subclass of Functor I'd have to include the arguments, and I have a ''lot'' of those subclasses! -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11523> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 33

[GHC] #11547: Accessing shadowed definitions in GHCi
by GHC 05 Jul '18

05 Jul '18

#11547: Accessing shadowed definitions in GHCi -------------------------------------+------------------------------------- Reporter: mniip | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: GHCi | Version: 8.0.1-rc1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Apparently GHCi creates a qualified namespace for each line it interprets, as can be observed with: {{{#!hs > data A = A > data A = B > :t A A :: Ghci3.A }}} Apparently accessing shadowed definitions in this way sometimes breaks GHCi: {{{#!hs > let a = a > let a = a > :t Ghci7.a <interactive>:1:1: error: • GHC internal error: ‘Ghci7.a’ is not in scope during type checking, but it passed the renamer tcl_env of environment: [] • In the expression: Ghci7.a }}} More specifically, complete reproduction instructions: {{{#!hs GHCi, version 8.1.20160205: http://www.haskell.org/ghc/ :? for help > let foo = foo > :t Ghci1.foo Ghci1.foo :: t > let foo = foo > :t Ghci1.foo <interactive>:1:1: error: • GHC internal error: ‘Ghci1.foo’ is not in scope during type checking, but it passed the renamer tcl_env of environment: [] • In the expression: Ghci1.foo }}} 7.10.2 (and 7.8.3 as verified over IRC by monochrom) appears to have a slightly different behavior. It seems it only creates namespaces for data/newtype/class declarations. Yet still, it fails with an error the first time but works completely fine thereafter: {{{#!hs > data A = A > data A = A > :t Ghci1.A Failed to load interface for ‘Ghci1’ no package key matching ‘interactive’ was found > :t Ghci1.A Ghci1.A :: Ghci1.A }}} 7.6.3 uses a different approach to displaying out of scope names so the behavior doesn't apply: {{{#!hs > data A = A > data A = B > :t A A :: main::Interactive.A }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11547> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #8263: allow duplicate deriving / standalone deriving
by GHC 04 Jul '18

04 Jul '18

#8263: allow duplicate deriving / standalone deriving ------------------------------------+------------------------------------- Reporter: aavogt | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.6.3 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+------------------------------------- I would like to avoid conditional compilation. Currently I need to write: {{{ #if MIN_VERSION_Cabal(1,18,0) -- cabal 1.18 has these instances #else deriving instance Typeable Flag #endif }}} Would it be sane to have the deriving instance do nothing if the imported instance that clashes was also derived? reifyInstances together with a fixed #8100 could allow faking this feature. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8263> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #10391: Ability to get contents of TH reified module
by GHC 03 Jul '18

03 Jul '18

#10391: Ability to get contents of TH reified module -------------------------------------+------------------------------------- Reporter: dmcclean | Owner: Type: feature | Status: new request | Milestone: Priority: low | Version: Component: Template | Operating System: Unknown/Multiple Haskell | Type of failure: None/Unknown Keywords: | Blocked By: Architecture: | Related Tickets: Unknown/Multiple | Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- After using reifyModule, I'd like to be able to inspect the list of exports from that module. At first I thought that this could perhaps be done by changing from {{{#!hs data ModuleInfo = -- | Contains the import list of the module. ModuleInfo [Module] }}} to {{{#!hs data ModuleInfo = -- | Contains the import list and export list of the module. ModuleInfo [Module] [Name] }}} but I now wonder if this is the best plan, because `thisModule >>= reifyModule` might be expected to function as a time machine, allowing us to see declarations that may not have been made yet (in the case where this module's export list is implicit). Can anyone propose a type that would allow us to perform this inspection for modules that have already been baked, but not for "this" module? An intended use case would be in our dimensional types library. We have several modules that define a whole bunch of units. It's useful for the unit name parser to have access to a dictionary with all of those units in it, but maintaining the definition of such a dictionary is redundant. If this feature existed, a module in the parser could import the modules where units are defined, examine the export lists, reify the names in the export lists, determine which represent unit definitions, and emit a declaration of a dictionary that contains them all. (This is in some ways similar to #9699, though perhaps more difficult?) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10391> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #12449: Broken types in identifiers bound by :print
by GHC 02 Jul '18

02 Jul '18

#12449: Broken types in identifiers bound by :print -------------------------------------+------------------------------------- Reporter: mniip | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: GHCi crash Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- {{{#!hs GHCi, version 8.0.1: http://www.haskell.org/ghc/ :? for help Prelude> :print show show = (_t1::Show a => a -> String) Prelude> :t _t1 <interactive>:1:1: error: No instance for (Show a) arising from a use of ‘it’ }}} Furthermore: {{{#!hs Prelude> _t1 "foo" <interactive>:3:5: error:ghc: panic! (the 'impossible' happened) (GHC version 8.0.1 for x86_64-unknown-linux): No skolem info: a_a1hz Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug Prelude> }}} Enabling `-fprint-explicit-foralls`, I think, reveals the issue: {{{#!hs Prelude> :set -fprint-explicit-foralls Prelude> :print id id = (_t2::a1 -> a1) Prelude> :t id id :: forall {a}. a -> a Prelude> :t _t2 _t2 :: a1 -> a1 }}} Similar behavior albeit with a different panic message happens on 7.8.4 and 7.10.3 as verified by bennofs over IRC: {{{ GHCi, version 7.10.3: http://www.haskell.org/ghc/ :? for help Prelude> :print print print = (_t1::Show a => a -> IO ()) Prelude> _t1 ghc: panic! (the 'impossible' happened) (GHC version 7.10.3 for x86_64-unknown-linux): tcTyVarDetails a_apc Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug }}} 7.6.3 doesn't exhibit this issue: {{{ GHCi, version 7.6.3: http://www.haskell.org/ghc/ :? for help Prelude> :print id id = (_t1::forall a. a -> a) }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12449> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #12410: Somehow detect splicing in ghci
by GHC 02 Jul '18

02 Jul '18

#12410: Somehow detect splicing in ghci -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- I'm no TH expert but currently it seems you need a hack (adding `data X ;` or `pure []`) to do the following {{{ ghci> $(reify ''() >>= runIO.print >> return []) <interactive>:210:3: error: • Couldn't match type ‘[t0]’ with ‘Exp’ Expected type: ExpQ Actual type: Q [t0] • In the expression: reify ''() >>= runIO . print >> return [] In the untyped splice: $(reify ''() >>= runIO . print >> return []) }}} {{{#!hs >>> data X; $(reify ''() >>= runIO.print >> return []) TyConI (DataD [] GHC.Tuple.() [] Nothing [NormalC GHC.Tuple.() []] []) >>> pure []; $(reify ''() >>= runIO.print >> return []) TyConI (DataD [] GHC.Tuple.() [] Nothing [NormalC GHC.Tuple.() []] []) }}} Same with [https://hackage.haskell.org/package/lens-4.14/docs/Control- Lens-TH.html#v:makeLenses makeLenses] (discussed [https://artyom.me/lens- over-tea-6 here], [https://github.com/ekmett/lens/issues/461 there]): {{{ >>> data A = B { _int :: Int } >>> makeLenses ''A <interactive>:209:1: error: • No instance for (Show DecsQ) arising from a use of ‘print’ • In a stmt of an interactive GHCi command: print it }}} the following two work {{{#!hs >>> data A = B { _int :: Int } >>> pure []; makeLenses ''A >>> view int (B 42) 42 }}} {{{#!hs >>> data A = B { _int :: Int }; makeLenses ''A >>> view int (B 42) 42 }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12410> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 10