[GHC] #14892: Field imposters with DuplicateRecordFields
#14892: Field imposters with DuplicateRecordFields -------------------------------------+------------------------------------- Reporter: philderbeast | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Keywords: | Operating System: MacOS X Architecture: x86_64 | Type of failure: GHC accepts (amd64) | invalid program Test Case: | Blocked By: Blocking: | Related Tickets: 13644 Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- There's a [[https://github.com/BlockScope/ghc-panic-12158|test case]] for this ghc panic. If I use `NamedFieldPuns` I wish this could more delicately pick the right field. If my record use is qualified then please restrict the set of candidate field names to use punned. {{{ {-# LANGUAGE DuplicateRecordFields #-} module Geodesy (X(..), Y(..)) where data X a = X {x :: a} data Y a = Y {x :: a} }}} {{{ {-# LANGUAGE NamedFieldPuns #-} module GhcPanic12158 where import qualified Geodesy as G (X(..)) import Geodesy (Y(..)) update :: G.X a -> G.X a update G.X{x} = G.X{x = x} }}} {{{
stack build ghc-panic-translateConPatVec-lookup-0.1.0: build (lib) Preprocessing library for ghc-panic-translateConPatVec-lookup-0.1.0.. Building library for ghc-panic-translateConPatVec-lookup-0.1.0.. [1 of 3] Compiling Geodesy [2 of 3] Compiling GhcPanic12158 ghc: panic! (the 'impossible' happened) (GHC version 8.2.2 for x86_64-apple-darwin): translateConPatVec: lookup
Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug }}} There are some fixes; 1. Add DuplicateRecordFields. {{{ {-# LANGUAGE NamedFieldPuns #-} {-# LANGUAGE DuplicateRecordFields #-} module GhcPanic12158 where import qualified Geodesy as G (X(..)) import Geodesy (Y(..)) update :: G.X a -> G.X a update G.X{x} = G.X{x = x} }}} 2. Use qualified field names. {{{ {-# LANGUAGE NamedFieldPuns #-} module GhcPanic12158 where import qualified Geodesy as G (X(..)) import Geodesy (Y(..)) update :: G.X a -> G.X a update G.X{G.x} = G.X{G.x = x} }}} Interestingly, if I don't import the record with the clashing field name then GHC complains. {{{ {-# LANGUAGE NamedFieldPuns #-} module GhcPanic12158 where import qualified Geodesy as G (X(..)) update :: G.X a -> G.X a update G.X{x} = G.X{x = x} }}} {{{
stack build ghc-panic-translateConPatVec-lookup-0.1.0: unregistering ghc-panic-translateConPatVec-lookup-0.1.0: build (lib) Preprocessing library for ghc-panic-translateConPatVec-lookup-0.1.0.. Building library for ghc-panic-translateConPatVec-lookup-0.1.0.. [3 of 3] Compiling GhcPanic12158
/ghc-panic-12158/earth/library/GhcPanic12158.hs:8:12: error: Not in scope: ‘x’ | 8 | update G.X{x} = G.X{x = x} | ^ /ghc-panic-12158/earth/library/GhcPanic12158.hs:8:21: error: Not in scope: ‘x’ | 8 | update G.X{x} = G.X{x = x} }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14892 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler
#14892: Field imposters with DuplicateRecordFields and NamedFieldPuns. -------------------------------------+------------------------------------- Reporter: philderbeast | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Resolution: | Keywords: Operating System: MacOS X | Architecture: x86_64 Type of failure: GHC accepts | (amd64) invalid program | Test Case: Blocked By: | Blocking: Related Tickets: 13644 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14892#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler
#14892: Field imposters with DuplicateRecordFields and NamedFieldPuns. -------------------------------------+------------------------------------- Reporter: philderbeast | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.4.1-alpha3 Resolution: | Keywords: ORF Operating System: MacOS X | Architecture: x86_64 Type of failure: GHC accepts | (amd64) invalid program | Test Case: Blocked By: | Blocking: Related Tickets: #13644 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by RyanGlScott): * cc: adamgundry (added) * keywords: => ORF * version: 8.2.2 => 8.4.1-alpha3 * related: 13644 => #13644 Comment: FWIW, this does not panic on GHC 8.4.1 (after #13644 was fixed), but instead gives an incorrect error message: {{{ $ /opt/ghc/8.4.1/bin/ghci GhcPanic12158.hs GHCi, version 8.4.0.20180224: http://www.haskell.org/ghc/ :? for help Loaded GHCi configuration from /home/rgscott/.ghci [1 of 2] Compiling Geodesy ( Geodesy.hs, interpreted ) [2 of 2] Compiling GhcPanic12158 ( GhcPanic12158.hs, interpreted ) GhcPanic12158.hs:8:12: error: • Constructor ‘G.X’ does not have field ‘x’ • In the pattern: G.X {x} In an equation for ‘update’: update G.X {x} = G.X {x = x} | 8 | update G.X{x} = G.X{x = x} | ^ }}} I believe this lines up with Adam's work on overloaded record fields, so I'll label this ticket as such. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14892#comment:2 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler
#14892: Field imposters with DuplicateRecordFields and NamedFieldPuns. -------------------------------------+------------------------------------- Reporter: philderbeast | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.4.1-alpha3 Resolution: | Keywords: ORF Operating System: MacOS X | Architecture: x86_64 Type of failure: GHC accepts | (amd64) invalid program | Test Case: Blocked By: | Blocking: Related Tickets: #13644 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by bgamari): Thanks for reducing this, philderbeast! -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14892#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler
#14892: Field imposters with DuplicateRecordFields and NamedFieldPuns. -------------------------------------+------------------------------------- Reporter: philderbeast | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.4.1-alpha3 Resolution: | Keywords: ORF Operating System: MacOS X | Architecture: x86_64 Type of failure: GHC accepts | (amd64) invalid program | Test Case: Blocked By: | Blocking: Related Tickets: #13644 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by philderbeast): You beat me to the test with a later GHC version, something Ben asked me to do on #12158. I too am seeing the problem of GHC's panic fixed. I do wish however that the compiler here would pick the one `x` pun that fits. {{{ # stack.yaml resolver: ghc-8.4.0.20180118 compiler: ghc-8.4.0.20180118 compiler-check: match-exact }}} {{{
stack build ghc-panic-translateConPatVec-lookup-0.1.0: build (lib) Preprocessing library for ghc-panic-translateConPatVec-lookup-0.1.0.. Building library for ghc-panic-translateConPatVec-lookup-0.1.0.. [2 of 3] Compiling GhcPanic12158
/earth/library/GhcPanic12158.hs:9:12: error: • Constructor ‘G.X’ does not have field ‘x’ • In the pattern: G.X {x} In an equation for ‘update’: update G.X {x} = G.X {x = x} | 9 | update G.X{x} = G.X{x = x} }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14892#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler
#14892: Field imposters with DuplicateRecordFields and NamedFieldPuns. -------------------------------------+------------------------------------- Reporter: philderbeast | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.4.1-alpha3 Resolution: | Keywords: ORF Operating System: MacOS X | Architecture: x86_64 Type of failure: GHC accepts | (amd64) invalid program | Test Case: Blocked By: | Blocking: Related Tickets: #13644 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by adamgundry): I think you want `DisambiguateRecordFields`. Without it, the 8.4 behaviour is correct, isn't it? Indeed `G.X` does not have a field `x` in scope (although it does have `G.x`). The error message could be better, perhaps suggesting `DisambiguateRecordFields`. Note that `DuplicateRecordFields` implies `DisambiguateRecordFields`, which is why enabling the former fixes the problem. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14892#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler
participants (1)
-
GHC