[darcs-users] Darcs on OS X (was: Can't record / Hashes mismatches)

Florian Gilcher darcs-lists at andersground.net
Tue Apr 27 07:52:30 UTC 2010

On Apr 26, 2010, at 8:30 PM, Eric Kow wrote:

> On Sat, Apr 17, 2010 at 18:27:17 +0200, Florian Gilcher wrote:
>> 2) I do have some experience with building installers under OS X, but
>> I am not sure how to build a universal binary for darcs on OS X. I
>> would volunteer to do the building, but I can't do the testing.
> Would you be willing to at least run the test suite?  (the other
> pre-requisite is us figuring out things out on a policy level, but
> getting a volunteer in the wings is nice).

No problem. There is one test failing. I will report that later (involves 
matching dates).

> I won't mind an Intel-only build as I suspect this will cover a large
> majority of cases.

Well, the big problem are Universal (i386/x86_64) binaries, which GHC is
still unable to generate (at least not that I am aware of). That extends: 
The Haskell platform is universal, but the cabal-installed libraries 
are not. So the only way to generate a universal darcs binary would be
to compile it with a i386 tool chain and a x86_64 tool chain and start 
playing with lipo.

Because of that, I would suggest supporting only i386 builds at the moment.
Most probably, users won't notice and Snow Leopard runs them just fine, even
in x86_64 mode.

Concerning darcs from MacPorts: it is broken (yet again), now because of 
hashed-storage still depending on mmap < 0.5:


Also, GHC takes long to build and you cannot use the Haskell platform provided
by haskell.org (as MacPorts insists on its own).

After all those considerations, I think it would be worthwhile to provide a
i386 static binary as well as an installer for darcs on OS X 10.5/10.6.

Florian Gilcher

More information about the darcs-users mailing list