xcode 5 psa info on haskell platform page

[Carter Schonwald]

Lets zoom out:

GHC HQ (with Darin's amazing help), are going to make available a 7.6.4
hotfix release in the next few weeks. Some last details are still underway
for that, but the GHC devs who support OS X seem to all be in agreement
that this matters and has to happen.

GHC 7.6.4 will be the recommended GHC 7.6 variant  for OS X folks, whether
they use haskell platform or not. (for those who don't choose to migrate
7.8 immediately, which will be many, for the usual reasons)

Haskell Platform Fall 2013 is coming soon, and we strongly recommend that
this fall's HP release use this new 7.6.4 GHC, because that will enable
there to be a single haskell platform release that will work on all OS X
variants without an additional burden on Haskell Platform.


Mark, GHC 7.6.4 will exist soon, we are advising you and the HP folks that
GHC HQ can't provide support for the OS X issues on 7.6 without doing a
7.6.4 release.

Speaking as someone who's done A LOT of support helping people deal with
Xcode 5 and OS 10.9 update woes in the past 4+ months, and recognizing that
there *will be people somehere using 10.8 / xcode 4 and then upgrading over
the next few months*, the wrapper approach will break there. Are you
committing to maintaining that wrapper?

much of the clang / gcc handling logic is already in head, so its mostly a
backport of that, and any gaps in the functionality need to be fixed up for
7.8 too!


On Mon, Oct 28, 2013 at 2:36 AM, Darin Morrison <darinmorrison at gmail.com>wrote:

> Mark,
>
> It seems we disagree about the fundamental issues. My stance is the
> following:
>
> 1) Current and future OS X releases of the HP should be buildable
> (including dependencies) on the current OS X release
>
> 2) We should avoid OS-specific hacks (like wrappers) and solve the
> underlying issues causing breakage
>
> And that's it.
>
> I understand that the HP does not re-build GHC and I understand that GHC,
> the HP, etc. need to work on multiple versions of OS X. But I also think
> it's possible to achieve portability in a cleaner fashion than relying on
> the proposed hacks.
>
> I don't know who makes the decisions about a 7.6.4 release. All I know is
> that there's been some discussion about it in private emails and some
> discussion on the #ghc IRC channel. As far as I understood, the primary
> impetus for such a release would be to provide a proper solution to GHC on
> 10.9.
>
> From where I'm sitting, if the current version of the HP is not buildable
> on the current OS X, it's simply broken. What should be fixed upstream will
> end up becoming (and in fact already is) a problem for others downstream.
> This just exacerbates the current mess that we're in because I can promise
> you that different users and different groups downstream will find and
> suggest a variety different solutions. Not only is that's bad for people
> that want things to "just work", it potentially introduces more headaches
> for everyone involved since certain approaches may create new problems that
> users will seek support for.
>
> You mentioned earlier that you wanted to avoid a package-manager based
> solution so as to remain neutral. I agree wholeheartedly with this stance
> but I also think being neutral entails that everything works for the
> largest group of people possible. Many of those people will either try to
> compile the HP themselves (on 10.9, without GCC easily accessible), or
> install through one of the package managers.
>
> Perhaps they are not the majority, but they are important nonetheless, and
> they include many developers from other communities who are used to
> development on OS X but not Haskell development specifically.
>
> I think that's about all I can say on this. I'm not involved in the
> official releases or maintenance of the HP stuff so it's not my call. I'm
> happy to help in trying to find a solution that works for everyone. I'm
> just not convinced the current proposal does that.
>
> Cheers,
> Darin
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://projects.haskell.org/pipermail/haskell-platform/attachments/20131028/efd02ec7/attachment.htm>