TH naming problem, different 'newName's clash

[Duncan Coutts]

I'm using the new abstract names of TH2 in ghc 6.3 (CVS early March)

I'm generating code that looks like this:

let foo_1 = e1
    foo_2 = e2
 in e3

I'm using 
name <- newName "foo"
to generate these names (foo_1, foo_2 etc) however when I splice this
code in ghc complains as if I'd written

let foo = e1
    foo = e2
 in e3

ie that I have multiple definitions of the same variable in the same let
block.

So, to be precise:

foo :: ExpQ
foo = do
  foo1 <- newName "foo"
  foo2 <- newName "foo"
  letE [valD (varP foo1) (normalB [| 3 |]) []
       ,valD (varP foo2) (normalB [| 4 |]) []]
       [| 5 |]

If we print this, it looks like what we want:
> do {e <- runQ foo; print (pprExp 0 e)}
let foo_0 = 3
    foo_1 = 4
 in 5

but if we ask about $(foo), ghc tells us:

    Conflicting definitions for `foo'
    In the binding group for: foo, foo

It looks like the unique identifiers on the local names are being
forgotten when we splice in.

I can work around this for the moment using something like
mkName ("foo_" ++ show somethingUnique)
but it's not nice.

Duncan