Oh, sorry, missed the first line. Building against GHC snapshots isn't supported. On Tue, Jun 7, 2011 at 6:48 AM, Don Stewart wrote:

It should build. If it doesn't, please report a bug.

On Tue, Jun 7, 2011 at 8:53 AM, Tim Docker wrote:

...

I'd like to build the haskell platform against a recent GHC snapshot, for testing purposes.

I see that I can download the source for the platform from:

http://lambda.galois.com/hp-tmp/2011.2.0.1/haskell-platform-2011.2.0.1.tar.g...

with instructions:

   http://www.vex.net/~trebla/haskell/haskell-platform.xhtml

However when I tried this some months ago, I found that some of the packages in the above source fail to build due to recent ghc changes (eg it's stricter handling of bang patterns).

Is there a more recent "beta" snapshot of the platform? Or a darcs/git repo somewhere? I imagine wanting to build the platform against bleeding edge ghc would be a pretty common desire.

Tim

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

On 7 June 2011 21:34, Tim Docker wrote:

Surely wanting to test against a ghc snapshot isn't that odd? How do others go about testing their code with many hackage dependencies against a new ghc? I would have expected that the first thing to do would be get the platform up and running.

Since the platform never uses the fist in a new major series of GHC, for non-boot libraries the process is that the first new major release of GHC comes out, code gets developed against that, then it's ready for the next release of the Platform. -- Ivan Lazar Miljenovic Ivan.Miljenovic@gmail.com IvanMiljenovic.wordpress.com

On 07/06/11 14:03, Brandon Allbery wrote:

On Tue, Jun 7, 2011 at 07:34, Tim Docker wrote:

...

do others go about testing their code with many hackage dependencies against a new ghc? I would have expected that the first thing to do would be get the

We don't, for the most part; ghc is a quickly moving target and usually has multiple type theory experiments going on in its code, and it's best to avoid new versions until they stabilize and any core dependents are updated. GHC announcements will often mention how usable the release is expected to be for us non-type theorists. Unless you're a ghc hacker, you shouldn't in general rush into a new ghc version. (This isn't Linux. The bleeding edge is for type theory, not geeks.)

I get your point, but it's not only type theory breakthroughs that are in the "bleeding edge". There's significant runtime improvements also: http://www.mail-archive.com/glasgow-haskell-users@haskell.org/msg19780.html (for certain bytestring usage patterns: >10x improvement in gc time, and a third as much memory used!). I was keen to try this out, but underestimated the effort in how much manual tweaking I'd have to do to get all the dependencies of my code working. Perhaps it's best for me to take Simon's advice and wait for the 7.2 release. Tim