Re: mingw regression test fixes

Claudio Natoli wrote:
> Peter, did you mean $host_platform by any chance... as originally
> submitted? I see no use of $build_platform anywhere, and as committed
> this plain fails.

I meant that $build_platform would be better than `uname`, not that my
word is to be taken blindly without testing.  But...

>
> I don't see why there is any need for discussion over this. There are
> 3 cases statements preceding this one, all of which use
> $host_platform. The one immediately preceding it does the very same
> thing for DIFFFLAGS!

... that is a very good argument for $host_platform as originally
proposed.