xcode 5 psa info on haskell platform page

[Darin Morrison]

Basically there were two problems:

1) 10.9/Xcode 5 don't ship with (or officially support) GCC and 7.6.3
doesn't currently compile with clang

2) 10.9 does not compile 7.6.3 even when the user installs GCC anyway.

Homebrew and Macports solve this by first installing GCC and (at least with
Homebrew—not sure about Macports) applying a small patch to fix the
remaining 10.9 problems.

I think we basically just need to put out a new GHC release (7.6.4) that
fixes things properly and use that for some new binaries.

I created a branch on my GHC github repo (
https://github.com/darinmorrison/ghc/tree/ghc-7.6.3-clang+mavericks) that
back-ports Austin's patches since the 7.6.3 release that make GHC
compilable with clang. It also includes the 10.9 fixes. (And should
probably include the fixes for dynamic linking too actually).

There's been some talk about putting together a 7.6.4 release next week. I
think that's the way to go. If not an official 7.6.4, there needs to be at
least something like a 7.6.3-hotfix-mavericks release.

I personally think relying on a wrapper is the wrong way to go in this. It
might work most of the time but it's a hack and hacks like that have a
tendency to cause other problems down the line.

We will already basically have to put out new packages even if we went with
the wrapper approach. Might as well take the opportunity to just distribute
fixed binaries.

On Saturday, October 26, 2013, Mark Lentczner wrote:

> Wait wait -
>
> Didn't I see a fix that allowed current GHC 7.6.3 to work on both
> Mavericks and Xcode 5. I believe this involved both a) using a clang
> wrapper, and b) re-writing the ghc settings file to set ghc's idea of gcc
> to the clang wrapper.
>
> This approach seems best to me, save one thing: The clang wrapper was
> written in Haskell. But it could easily be a small shell script.
>
> I'd like to avoid any solution that involves brew or macports: Not all
> developers have those on their system, and often putting them on your
> system requires "buying into" their approach, and so I don't want to
> inflict it on anyone (least of all myself - I won't run them - mostly to
> ensure I have an 'agnostic' machine.)
>
> Thoughts? I can putter with this tonight.
>
> - Mark
>
>

-- 
Darin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://projects.haskell.org/pipermail/haskell-platform/attachments/20131026/e7146262/attachment.htm>