xcode 5 psa info on haskell platform page

[Darin Morrison]

On Sun, Oct 27, 2013 at 1:18 PM, Mark Lentczner <mark.lentczner at gmail.com>wrote:

> I've reviewed all the changes in that branch - and they seem to fall into
> two categories:
>
> 1) Changes the flags that ghc passes to whatever it thinks is gcc, so that
> they work with clang.
> 2) Changes to ghc source so that clang can compile it.
>

Not quite. Two of the patches (the ones authored by me) are needed to build
7.6.3 on 10.9 regardless of whether clang or GCC are used.


> For the platform, we don't care about #2 - we don't ship the GHC source,
> and we don't expect platform users to build it.
>

I'm not suggesting that we change the principles of the Haskell Platform to
expect users to build their own binaries.

However, there are a fair number of users who do this—either directly or
indirectly (via Homebrew or Macports). In general, I think it's a bit of a
problem that it's not currently possible to build GHC and the rest of the
Haskell Platform on 10.9.

I think it would be especially bad to ship a _new_ Haskell Platform release
that can't be built on the current version of OS X.


> Items in #1 look to my eye like they are covered by the wrapper script (or
> some equivalent that we can build with it). What's more, the changes in #1
> seem to rely on the idea that GHC was built knowing whether it will be used
> with gcc or with clang. This seems undesirable to me (at least for now),
> insofar as there will be people running Xcode 4 for the foreseeable future,
> and I'd really rather not have to produce variants of GHC or HP just to
> support Xcode 4 vs. 5.
>

I'm not sure if this is accurate, but it is a problem if correct. There's
also the issue of C compiler command being hard-coded into the settings
file.


> I don't see anything in those changes that handles the fact that gcc is
> hard coded into hsc2hs, though I might be misunderstanding that issue.
>
> So - looks to me like a bash script wrapper, and redirecting ghc's
> settings is still the best option.
>

I still don't like the idea of a wrapper script, especially if it's a hack
specific only to the OS X version.

I would prefer depending on an updated version of GHC which 1) compiles
cleanly on 10.9, 2) works with clang and GCC, and 3) either dynamically
detects which C compiler is available or offers a static configuration
method like xcode-select but which updates the GHC settings. Something like
version (3) could be run during install time, preventing the need for
separate binaries.

That's just my take on it. I realize it would require more work but I think
it would be a cleaner and more reliable approach than a wrapper script.

—Darin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://projects.haskell.org/pipermail/haskell-platform/attachments/20131027/db9fd44c/attachment-0001.htm>