Re: Failures on NetBSD/macppc 6.0 for pkgsrc-current

[Havard Eidnes <he%NetBSD.org@localhost>]

>>> Hmm, apparently http://gnats.netbsd.org/50037 hasn't been pulled up to
>>> -6 at all.
>>
>> It's not pulled up to be part of the 6.1 release, but the significant
>> part of it appears to have been pulled up to netbsd-6, [...]
>
> What does the "significant part" mean? Is the FPU-context switching bug
> fixed in netbsd-6 or not?

Sorry for being a bit vague.

The above PR was for netbsd-6 tackled by

  http://releng.netbsd.org/cgi-bin/req-6.cgi?show=1310

Initially, this set of changes were requested pulled up:

sys/arch/powerpc/include/psl.h			1.20
sys/arch/powerpc/include/userret.h		1.29, 1.30
sys/arch/powerpc/oea/altivec.c			1.30
sys/arch/powerpc/oea/oea_machdep.c		1.72
sys/arch/powerpc/powerpc/fpu.c			1.36

However, this resulted in build failures on netbsd-6, and was
subsequently reverted, and later on the following set of
revisions/patches were pulled up:

sys/arch/powerpc/include/userret.h              1.29-1.30 (patch)
sys/arch/powerpc/oea/oea_machdep.c              1.72 (patch)
sys/arch/powerpc/include/psl.h                  1.20 (patch)

with this text:

	Add a check to make sure that if PSL_FP is set, we own the FPU.
	Don't clear PSL_FP/PSL_VEC.
	Remove PSL_SPV from BOOKE PSL_USERMOD
	Don't preserve PSL_FP/PSL_VEC in cpu_setmcontext.
	[matt, ticket #1310]

It's this latter set of files I've updated via "cvs update -r netbsd-6"
and I'm otherwise using 6.1 release sources.

> It is fixed in netbsd-7, but 7 introduces some new instabilities for
> oea-platforms (ofppc, macppc, sandpoint, etc.) which are even worse, so I
> already downgraded some machines (especially servers) to 6 again. Would be
> nice to have an official netbsd-6 release with the fix, but without the new
> bugs. ;)

The fixes should be part of 6.2 when/if we release it.

Regards,

- Håvard