[committed] Skip floating-point var-expand tests for soft-float MIPS targets

[Hans-Peter Nilsson]

On Wed, 28 Nov 2007, Richard Sandiford wrote:
> For the record, I think this is one case where target-specific
> preprocessor tests are better.

But those preprocessor macros don't exist in the general case!

>  Grepping something as ill-defined
> as an RTL dump for a feature test sounds like something that could
> silently go awry for non-obvious reasons, and suddenly you're not
> running hard-float tests when you expected to.

*If* that ever happens, for those targets add overrides, testing
the preprocessor macro.  I argue that looking for '(call'
here *is* the cleaner, more general method *in this case*.

>  Using a preprocessor
> test also encourages people to add an appropriate built-in macro for
> users to test as well.

I don't see why this should be encouraged, for a target that
e.g. only has float-soft.

brgds, H-P