On 20/01/2011 11:57, Manuel M T Chakravarty wrote:
Roman Leshchinskiy:
Manuel M T Chakravarty wrote:
Roman Leshchinskiy:
Manuel M T Chakravarty wrote:
I would like that, too. (In fact, it is what I implemented at
first.) Unfortunately, it doesn't work. Language options are dynamic
options, but -package is a weird intermediate between static and
dynamic. Hence, if you put ParallelArrays (as it is now called) into
a LANGUAGE pragma, you won't get -fdph-par.
But the -fdph-* flags themselves are dynamic. Is it the case that
putting {-# OPTIONS_GHC -fdph-par #-} into a module works but having {-#
LANGUAGE
ParallelArrays #-} imply it doesn't? That would be weird.
IIRC, {-# OPTIONS_GHC -fdph-par #-} doesn't work either. GHC doesn't
complain, but it also doesn't make the package available.
Urgh, that's bad! Perhaps fixing this is worth a try? Ian, Simon, how hard
would it be to make -package properly dynamic?
Very hard as far as I could see.
I haven't really thought about it. In principle it shouldn't be *that*
hard, because in GHCi you can say ":set -package foo" and it rebuilds
the internal package database.
I'm not sure that {-# OPTIONS_GHC -package foo #-} is something we want
to be encouraging, but that's a separate question.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
