> Me too. GHC seems to generate wrong machine code when building itself.
> I can add the following information:
>
> When building ghc from source, using a ghc-6.2.1 (recent
> source tarball
> built with ghc-5.02.3 and installed), the first ghc which is
> built (stage 1)
> *seems* to work correctly. It builds the stage 2 ghc without
> showing any errors.
> Problems show up when using this ghc (stage 2), e.g. for building a
> stage 3 ghc, but also any trivial "hello world" program.
>
> Compiling with "ghc-inplace -v X.hs" shows that the compiler
> segfaults during
> the type check.
We're not seeing this (obviously, since the nightly builds are still
going through). Hmmm. It's not a gcc issue, because someone using gcc
3.2.2 also reported it, and that's what we use here.
What's your exact build setup: build.mk, any configure arguments?
Cheers,
Simon
|
