still building soft-fp on ppclinux

[Roland McGrath]

> As my original post specifies, yes I tested with and without fpu, and
> I verified that the resulting code had fpu code when it should (and
> the reverse).  I also checked that an fpu build did not have
> superfluous objects being generated (__addsf3.o, etc).

That is clearly not true of the code that actually got checked in.  Looking
at your patch in the mailing list archive, I don't see anything different
that would explain how your patches could possibly have worked with a
normal fpu-using powerpc32 build.  Either you are mistaken, or you did not
test with the current code base, or Ulrich managed to break your changes in
some way I am overlooking.  From what I can see they were broken to start.