Teach gimple_canonical_types_compatible_p about incomplete types

[Joseph Myers]

On Sat, 30 May 2015, Jan Hubicka wrote:

> Joseph, does the attached testcase make sense for you? Is it defined? It is my
> first attempt to really interpret C standard to detail.

I suppose it's defined if unsigned int is the type chosen as compatible 
with that enum.  The test should be skipped for short_enums targets 
(arm-eabi bare metal) (you can't simply use -fno-short-enums as then that 
will fail the link-time compatibility checking).

-- 
Joseph S. Myers
joseph@codesourcery.com