Re: PR27599

[Shantonu Sen <ssen at opendarwin dot org>]

I tested the following, with a C-only cross compiler from linux ->  
darwin

1) GCC trunk. Works
2) trunk + Michael's first patch, which exposes the latent bug. The  
build didn't fail, but gcc/Makefile had "xmake_file= $(srcdir)/config/ 
x-darwin $(srcdir)/config/x-linux"m which is wrong and may result in  
bugs in the future.
3) trunk + first patch + patch below. Works, and now I have  
"xmake_file= $(srcdir)/config/x-linux"

Shantonu

On May 16, 2006, at 7:03 AM, Michael Matz wrote:

> Hi,
>
> On Mon, 15 May 2006, Alexandre Oliva wrote:
>
> > On May 15, 2006, Michael Matz <matz@suse.de> wrote:
> >
> > > 	* config.host: Make assignments to host_xmake_file cumulative.
> >
> > Ok, thanks
>
> ssen@opendarwin.org mentioned that linux -> darwin cross compiler  
> would
> break with the patch because config.host contains a hidden bug since  
> ever,
> it currently contains:
>
> case ${target} in
>   *-darwin*)
>     # Generic darwin host support.
>     out_host_hook_obj=host-darwin.o
>     host_xmake_file="${host_xmake_file} x-darwin"
>     ;;
> esac
>
> I.e. selecting on $target.  This seems wrong to me, but I can't test  
> any
> change there.  Can I replace it with testing for $host, ala the patch
> below?
>
>
> Ciao,
> Michael.
>
> 	* config.host (*-darwin*): Test $host, not $target.
>
> --- config.host (revision 113823)
> +++ config.host (working copy)
> @@ -86,11 +86,11 @@ case ${host} in
>  esac
>
>  # Common parts for widely ported systems.
> -case ${target} in
> +case ${host} in
>    *-darwin*)
>      # Generic darwin host support.
>      out_host_hook_obj=host-darwin.o
> -    host_xmake_file=x-darwin
> +    host_xmake_file="${host_xmake_file} x-darwin"
>      ;;
>  esac