Splicing in constructors that are abstract from the point of view of the splice point is unreliable
Test1.hs:
{-# LANGUAGE TemplateHaskell #-}
module Main where
import Test2
main = putStrLn (show $(makeOne))
Test2.hs:
{-# LANGUAGE TemplateHaskell #-}
module Test2(ExportedAbstract, makeOne) where
data ExportedAbstract = Yay String | NonYay AnotherThing
data AnotherThing = Foo String
instance Show ExportedAbstract where
show _ = "Shown"
makeOne = [| Yay "Yep" |]
Compile this with optimizations:
ghc Test2.hs -O2
ghc --make Test1.hs
Results in:
$ ./Test1
Shown
Compile this without optimizations:
ghc Test2.hs -O0
ghc --make Test1.hs
Results in:
$ ghc --make Test1.hs
[2 of 2] Compiling Main ( Test1.hs, Test1.o )
Loading package base ... linking ... done.
Loading package pretty-1.0.0.0 ... linking ... done.
Loading package array-0.1.0.0 ... linking ... done.
Loading package packedstring-0.1.0.0 ... linking ... done.
Loading package containers-0.1.0.1 ... linking ... done.
Loading package template-haskell ... linking ... done.
Test1.hs:7:24:
Can't find interface-file declaration for data constructor Test2.Yay
Probable cause: bug in .hi-boot file, or inconsistent .hi file
Use -ddump-if-trace to get an idea of which file caused the error
In the first argument of `show', namely `$makeOne'
In the first argument of `putStrLn', namely `(show ($makeOne))'
In the expression: putStrLn (show ($makeOne))
Why is this? Because trimThing in TidyPgm makes data types abstract in some circumstances, and it only happens if compiling a boot file or with -O0!
I think this program should compile as the splicing is not a violation of the abstractness of the data type: all the manipulation etc is still being done by the defining module.
Trac metadata
Trac field | Value |
---|---|
Version | 6.8.3 |
Type | Bug |
TypeOfFailure | OtherFailure |
Priority | normal |
Resolution | Unresolved |
Component | Template Haskell |
Test case | |
Differential revisions | |
BlockedBy | |
Related | |
Blocking | |
CC | |
Operating system | |
Architecture |