Re: Recent changes causing atf test failures

[Ryota Ozaki <ozaki-r%NetBSD.org@localhost>]

On Thu, Jul 3, 2014 at 11:25 PM, Andreas Gustafsson <gson%gson.org@localhost> 
wrote:
> Greg Troxel wrote:
>> The releng tests aren't finishing either:
>>
>>   http://releng.netbsd.org/b5reports/i386/commits-2014.07.html#end
>
> The i386 runs aren't finishing, but the amd64 ones are.
> Before ozaki-r's commits, there were 29 failures:
>
>   http://releng.netbsd.org/b5reports/amd64/build/2014.07.02.00.15.41/test.html
>
> And after the commits, there are 88 failures:
>
>   http://releng.netbsd.org/b5reports/amd64/build/2014.07.02.09.51.08/test.html
>
> The fs/nfs/t_rquotad tests are timing out, and many other NFS related
> tests are failing with
>
>   panic: kernel diagnostic assertion "curcpu() == ci" failed

Thank you for your investigation. The rump kernel seems to be
compiled unexpectedly with the code supposed to be for normal kernels.
ifdef added by me in the commit
  http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/net/pktqueue.c.diff?r1=1.6&r2=1.7
was wrong. Looking at sys/rump/Makefile.rump, _RUMPKERNEL is defined
for most architectures, however, not for x86.

I'll confirm my guess is correct and fix it if so.

Thanks,
  ozaki-r

>
> --
> Andreas Gustafsson, gson%gson.org@localhost