32-bit host pdp11 breakage

[Stephen Casner]

On Thu, 3 Sep 2020, Alan Modra wrote:
> On Wed, Sep 02, 2020 at 10:05:07AM -0700, Stephen Casner wrote:
> > Does this complaint mean that the assembly language is constrained
> > not to use .quad for any target when the host is 32-bit?
>
> No, .quad should still work (and the .quad tests pass on pdp11).
> O_big rather than O_constant will be used for numbers that don't fit
> in a bfd_vma.

Let me make sure I'm interpreting your message correctly.  I get it
that for a 32-bit host md_number_to_chars won't be called with n=8
because O_big rather than O_constant will be used.  Does that mean
that the change you've made is a complete fix and you've already
verified that the .quad tests _do_ pass for pdp11?

Or does your patch just avoid the compiler error and there is still
more work that I need to do for a 32-bit host to process the .quad
directive correctly for pdp11?

> Incidentally, on x86_64 you can do a 32-bit host test by configuring
> with
> CFLAGS="-m32 -g -O2" CXXFLAGS="-m32 -g -O2" .../configure ...

Good point, but macOS doesn't like 32-bit code any more.  I accept
that this may be an unreasonable limitation.

                                                        -- Steve