Hello!
GHC on powerpc arch is built unregisterized, without interpreter, without native codegen. The result is stage-1 compiler which is unable to build Template Haskell programs. I've built GHC without
GhcUnregisterised=YES GhcWithNativeCodeGen=NO GhcWithInterpreter=NO SplitObjs=NO
Now it seems to work fine. So I'm asking to exclude option listed above on powerpc.
Hi Max,
On Tue, Mar 13, 2007 at 06:36:59PM +0300, Max Vasin wrote:
GHC on powerpc arch is built unregisterized, without interpreter,
without native codegen. The result is stage-1 compiler which is unable to build Template Haskell programs. I've built GHC without
GhcUnregisterised=YES GhcWithNativeCodeGen=NO GhcWithInterpreter=NO SplitObjs=NO
Now it seems to work fine. So I'm asking to exclude option listed above on powerpc.
While it may work today, the problem is that it has a tendency to break as gcc et al change, for example http://hackage.haskell.org/trac/ghc/ticket/1167
If it was just a matter of patching GHC and uploading it then it wouldn't be too much of a problem, but even if the new GHC is fixed the /current/ GHC will still fail to compile it. Thus it means either finding someone with a toolchain on powerpc that still works with GHC, or doing a bootstrap again, both of which are time consuming.
While this could be a problem on x86/amd64 too, those arches are actively developed and probably higher quality in the first place. Also, I can fix this on my own machines, which means much less overhead, as well as normally faster machines, no lag, etc.
I do have some patches to make ghci work on unregisterised ports, which I'll hopefully be able to whip into shape after 6.6.1 is sorted.
Thanks Ian
debian-haskell@lists.urchin.earth.li