add mips long_call attribute

[Richard Sandiford]

James E Wilson <wilson@specifix.com> writes:
> I'll admit that doing the long_call attribute check directly in
> const_call_insn_operand is a reasonable way to implement this.  I found
> at least 3 different ways to implement this feature, and I picked one
> that I thought was most reasonable.  It isn't necessary the same one
> that you will think is most reasonable.

Fair enough.  I certainly didn't want to make a big fuss about this.
I agree there's more than one way of coding it, and it just happened
that what seemed the most natural to you didn't seem the most natural
to me, so I was interested in the rationale.  Your replies have
certainly provided that, thanks.

So, no objection from me.

Richard