Re: More toolchain issues

[Frank Wille <frank%phoenix.owl.de@localhost>]

Martin Husemann wrote:

> On Tue, Sep 29, 2009 at 03:04:35PM +0200, Frank Wille wrote:
>> We should check the M68k/V.4-ABI specs, but I guess that the ABI defines
>> d0, d1, a0 and a1 as volatile registers over a function call. Integers
>> are returned in d0, (64-bit in d0/d1 ?) and pointers in a0.
> 
> Yes, we should check. gcc considers (CALL_USED_REGISTERS) d0/d1/a0/a1/a7
> as clobbered by a function call.

Right.


> And be sure to run all regress/ and test/ regression tests with the
> change.

Yes, I will do so. But I wouldn't dare to change anything without a comment
from mlelstv, mhitch or any other m68k port maintainer...

So, if there are any objections, please speak now! :)

-- 
Frank Wille