hGetContents: invalid argument (invalid byte sequence)

[hvr]

Currently, Language.Haskell.Extract, interacts with the LANG environment variable, leading to surprising failures. E.g. when compiling

{-# LANGUAGE UnicodeSyntax, TemplateHaskell #-}

module Foo where

import Language.Haskell.Extract

foo ∷ α → α
foo x = x

bar = $(functionExtractor "oo$")

without any set LANG variable on Linux, you get a

Foo.hs:10:9:
    Exception when trying to run compile-time code:
      Foo.hs: hGetContents: invalid argument (invalid byte sequence)
      Code: functionExtractor "oo$"
    In the expression: $(functionExtractor "oo$")
    In an equation for `bar': bar = $(functionExtractor "oo$")

This is caused by relying on System.IO.readFile to infer the text-encoding, instead of explicitly setting it to utf8.
Since GHC already enforces utf8 as source file encoding, Language.Haskell.Extract should follow GHC's behaviour (at least when compiled for GHC).