[jhc] darcs patch: Explicitly use base-4 and syb package. (GHC 6.10 only?)

[Corey O'Connor]

On Mon, Jun 22, 2009 at 10:56 AM, John Meacham<john at repetae.net> wrote:
> Cool. thanks for the patches!
>
> Yeah, I am trying to maintain both ghc-6.8.x and 6.10.x compatibility.
> however, I don't actually have a 6.10 system to test on so I have to
> rely on bug reports. From what I hear, it is mainly the packages that
> differ, so perhaps we can just maintain two PACKAGES lines in the
> Makefile for now and comment out the one we don't need.

Makes sense! I wasn't sure how to best make the PACKAGES parameter
depend on the ghc version. I'm an automake newbie though I presume
there is a way to have the value depend on a configuration value.

> What was the specific thing that relied on base 4 out of curiosity?

I'll unpull the patch to figure out exactly. However i think it was
just that if hide-all-packages is used then even the base package is
hidden. So any import of a module from base has to come from somewhere
and I just decided to use base-4 instead of base-3.

> I want to keep 6.8.2  around as that is what comes with the
> distributions used for amazon ec2 instances. firing up a high-cpu
> instance for regression testing of jhc is quite nice.

I have an AMI that has GHC 6.10.3 and cabal-install already setup up.
Not sure how to make it distributable but I'll let you know once I do.

> For the curious, right now I am working on better (as in fully
> transparent) cross-compilation support for jhc. there has been
> signifigant interest in using jhc to compile apps for things like
> iPhones. So you can expect something in a few days on that front.

Sweet! In addition to some other nice features of JHC, that's what
peeked my interest in JHC.

On another topic: Currently my build of JHC fails when trying to build
the containers library due to missing functionality in the
Data.Typeable module. Is this a known issue?

Cheers,
Corey O'Connor