Marge Bot pushed to branch master at Glasgow Haskell Compiler / GHC Commits: 3bd7dd44 by mangoiv at 2025-12-04T04:36:45-05:00 Renamer: reinstate the template haskell level check in notFound Out-of-scope names might be caused by a staging error, as is explained by Note [Out of scope might be a staging error] in GHC.Tc.Utils.Env.hs. This logic was assumed to be dead code after 217caad1 and has thus been removed. This commit reintroduces it and thus fixes issue #26099. - - - - - 4 changed files: - compiler/GHC/Tc/Utils/Env.hs - + testsuite/tests/th/T26099.hs - + testsuite/tests/th/T26099.stderr - testsuite/tests/th/all.T Changes: ===================================== compiler/GHC/Tc/Utils/Env.hs ===================================== @@ -8,6 +8,7 @@ -- in module Language.Haskell.Syntax.Extension {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE LambdaCase #-} +{-# LANGUAGE MultiWayIf #-} module GHC.Tc.Utils.Env( TyThing(..), TcTyThing(..), TcId, @@ -1213,6 +1214,20 @@ pprBinders bndrs = pprWithCommas ppr bndrs notFound :: Name -> TcM TyThing notFound name = do { lcl_env <- getLclEnv + ; lvls <- getCurrentAndBindLevel name + ; if -- See Note [Out of scope might be a staging error] + | isUnboundName name -> failM -- If the name really isn't in scope + -- don't report it again (#11941) + -- the + -- the 'Nothing' case of 'getCurrentAndBindLevel' + -- currently means 'isUnboundName' but to avoid + -- introducing bugs after a refactoring of that + -- function, we check this completely independently + -- before scrutinizing lvls + | Just (_top_lvl_flag, bind_lvls, lvl@Splice {}) <- lvls + -> failWithTc (TcRnBadlyLevelled (LevelCheckSplice name Nothing) bind_lvls (thLevelIndex lvl) Nothing ErrorWithoutFlag) + | otherwise -> pure () + ; if isTermVarOrFieldNameSpace (nameNameSpace name) then -- This code path is only reachable with RequiredTypeArguments enabled @@ -1243,14 +1258,23 @@ wrongThingErr expected thing name = {- Note [Out of scope might be a staging error] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Consider - x = 3 - data T = MkT $(foo x) + type T = Int + foo = $(1 :: T) + +GHC currently leaves the user some liberty when it comes to using +types in a manner that is theoretically not well-staged. +E.g. if `T` here were to be a value, we would reject the program with +a staging error. Since it is a type though, we allow it for backwards +compatibility reasons. + +However, in this case, we're just in the process of renaming a splice +when trying to type check an expression involving a type, that hasn't +even been added to the (type checking) environment yet. That is, why +it is out of scope. -where 'foo' is imported from somewhere. +The reason why we cannot recognise this issue earlier is, that if we +are not actually type checking the splice, i.e. if we're only using the +name of the type (e.g. ''T), the program should be accepted. -This is really a staging error, because we can't run code involving 'x'. -But in fact the type checker processes types first, so 'x' won't even be -in the type envt when we look for it in $(foo x). So inside splices we -report something missing from the type env as a staging error. -See #5752 and #5795. +We stop and report a staging error. -} ===================================== testsuite/tests/th/T26099.hs ===================================== @@ -0,0 +1,6 @@ +{-# LANGUAGE TemplateHaskell #-} +module M where + +type T = Int + +a = $(3 :: T) ===================================== testsuite/tests/th/T26099.stderr ===================================== @@ -0,0 +1,6 @@ +T26099.hs:6:12: error: [GHC-28914] + • Level error: ‘T’ is bound at level 0 but used at level -1 + • In an expression type signature: T + In the expression: 3 :: T + In the untyped splice: $(3 :: T) + ===================================== testsuite/tests/th/all.T ===================================== @@ -642,3 +642,4 @@ test('QQInQuote', normal, compile, ['']) test('QQTopError', normal, compile_fail, ['-fdiagnostics-show-caret']) test('GadtConSigs_th_pprint1', normal, compile, ['']) test('GadtConSigs_th_dump1', normal, compile, ['-v0 -ddump-splices -dsuppress-uniques']) +test('T26099', normal, compile_fail, ['']) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/3bd7dd44152f74d7a9fdd036f26be10c... -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/3bd7dd44152f74d7a9fdd036f26be10c... You're receiving this email because of your account on gitlab.haskell.org.