Aymeric Vincent <vincent@...> writes:
> Juho Snellman <jsnell@...> writes:
>> Cool. I'm slightly surprised at this, since the earlier attempt at an
>> NetBSD/ppc port by Richard Kreuter ran into trouble with 64-bit values
>> in the FFI. (see http://thread.gmane.org/gmane.lisp.steel-bank.devel/6074/)
> Hm, I wanted to investigate that, but I'm quite busy at the
> moment... The only simple explanation that comes to mind is that I
> based my patch on the Linux version, and both Linux and NetBSD use the
> same ELF ABI so there is no reason why 64-bit values would work on
> one and not on the other. Richard's patch, as far as I saw, was rather
> based on the Darwin port, and this could be the reason for the
> differences. I'm not familiar enough with Darwin to tell.
It might be that 64-bit ppc FFI support was reworked in the time since
Richard last tried to port it?
>> > [...]
>> It would probably be better to keep that as #!+linux, and have a
>> separate space layout block for NetBSD. Even if the spaces happen to
>> be the same now, I'd prefer that the Linux layout can be modified
>> independently from the NetBSD layout in the future.
> Alright, I changed this and recompiled everything. This time I attach
> the patch.
Thank you. I merged it, with some minor changes, as sbcl-220.127.116.11. I
don't have a NetBSD/PPC machine, so I haven't tested it there; I did
test it on Darwin/PPC, and the tests mostly passed. If someone on
this list has an easy way to test CVS on a Linux/PPC machine, that
would be appreciated.
(On Darwin/PPC, a debug.impure.lisp test marked :BUG-346 failed. I
can't imagine that this was caused by your patch; however, it would be
good to know whether that is currently failing only on Darwin or
whether NetBSD and Linux are also failing.)
Many thanks again,