[kaffe] mipsel and jit3

[Dalibor Topic]

Casey Marshall wrote:
>>>>>>"Dalibor" == Dalibor Topic <robilad at kaffe.org> writes:
> 
> 
> Dalibor> Mikolaj Habryn wrote:
> 
>>>On Tue, 2004-10-12 at 14:10, Dalibor Topic wrote:
>>>
>>>
>>>>Sounds like a good guess. Could you reconfigure and try again
>>>>using libffi?
>>>
>>>Didn't change the situation, alas. I'm faintly suspicious of
>>>libffi; I had thought that it didn't support MIPS (and the MIPS
>>>habit of changing the calling convention for leaf vs non-leaf
>>>functions makes me wonder how libffi manages at all!) but I'm still
>>>poking at it.
> 
> 
> Dalibor> Hi Mikolaj,
> 
> Dalibor> I've CC:ed Casey, who's been hacking on libffi for MIPS
> Dalibor> recently. You may want to drag out libffi straight from gcc
> Dalibor> CVS, as no gcc release contains Casey's fixes yet afaik.
> 
> Yeah, my (and David Daney's) fixes are in GCC CVS right now, but note
> that it probably only has support (at least, support for closures) on
> the O32 ABI. I don't know how well calling works on N32 or N64, but it
> is presumably ok.
> 
> Does Kaffe use calls only, or closures? If it is only calls, then MIPS
> support should be pretty good for any ABI.

After consulting with Casey on IRC, I can say that Kaffe doesn't use 
libffi closures. So libffi from gcc CVS might be a rather good choice of 
a libffi.

cheers,
dalibor topic