This is a bug in Hugs.
Hugs doesn't try to remember what entities it brought into scope from which
module so when you write:
module Foo( module Bar ) where
it exports everything exported from module Bar instead of everything that Foo
imported from module Bar.
It's much simpler to implement and easier to understand - but not what we're
supposed to be doing.
I'm adding it to the known bugs list but not planning to try fixing this
anytime soon (unless it's really hurting someone).
Alastair
> Section 5.1.1 of the Haskell 1.4 report says:
>
> The set of all entities brought into scope from a module m by one or
> more unqualified import declarations may be named by the form `module
> m', which is equivalent to listing all of the entities imported from
> the module.
> I'm getting strange behavior from both GHC and Hugs w.r.t. module
> exportations. They disagree with each other somewhat and both seem wrong,
> although I'm not certain I understand the report on this matter.
>
> Here are two test programs. First Test1.hs:
>
> module Test1 (module Test1) where
> main = putStrLn "Test1's main"
> foo = "Test1 foo"
>
> In Test2.hs, I want to modify and extend Test1, keeping "foo" but
> replacing main.
>
> module Test2 (module Test1, module Test2) where
> import Test1 hiding (main)
> main = putStrLn "Test2's main"
> bar = foo ++ " plus Test2 bar"
>
> I also have test programs TMain1.hs:
>
> module Main where
> import qualified Test1 as Test
> main = Test.main
>
> and TMain2.hs:
>
> module Main where
> import qualified Test2 as Test
> main = Test.main
>
> and makefile rules:
>
> test1.exe : Test1.o TMain1.o
> test2.exe : Test2.o Test1.o TMain2.o
>
> Compiling and running these guys under GHC, I get:
>
> c:\Users\conal\Fran\demos\CurveEditor>test1
> test1
> Test1's main
>
> c:\Users\conal\Fran\demos\CurveEditor>test2
> test2
> Test1's main
>
> Under Hugs, I get this same behavior
>
> c:\Users\conal\Fran\demos\CurveEditor>runhugs TMain1
> Test1's main
>
> c:\Users\conal\Fran\demos\CurveEditor>runhugs TMain2
> Test1's main
>
> Swapping the export order in Test2 gives me what I want in Hugs:
>
> c:\Users\conal\Fran\demos\CurveEditor>runhugs TMain2
> Test2's main
>
> but no change in GHC.
>
> c:\Users\conal\Fran\demos\CurveEditor>make test2.exe
> rm -f Test2.o
> /fptools/bin/i386-unknown-cygwin32/ghc-2.09/ghc -fglasgow-exts
> -concurrent -recomp -cpp
> -i../../src:../../src/GHC:../../gc/GHC:/fptools/lib/i386-unknown-cygwin32/gh
> c-2.09/win32 -c Test2.hs -o Test2.o -ohi Test2.hi
> Module version unchanged at 3
> /fptools/bin/i386-unknown-cygwin32/ghc-2.09/ghc -fglasgow-exts
> -concurrent -recomp -cpp
> -i../../src:../../src/GHC:../../gc/GHC:/fptools/lib/i386-unknown-cygwin32/gh
> c-2.09/win32 -optl-u -optl_NoRunnableThreadsHook -o test2 \
> Test2.o Test1.o TMain2.o -L../../src
> -L/fptools/lib/i386-unknown-cygwin32/ghc-2.09/win32 -lFran -lWin32
> -L../../SpriteLib -lSpriteLib -luser32 -lgdi32
>
> c:\Users\conal\Fran\demos\CurveEditor>test2
> Test1's main
>
> Section 5.1.1 of the Haskell 1.4 report says:
>
> The set of all entities brought into scope from a module m by one or
> more unqualified import declarations may be named by the form `module
> m', which is equivalent to listing all of the entities imported from
> the module.
>
> - Conal
