Re: [linux-linus test] 166151: regressions - FAIL

[Ian Jackson <iwj@xxxxxxxxxxxxxx>]

Jan Beulich writes ("Re: [linux-linus test] 166151: regressions - FAIL"):
> On 16.11.2021 05:42, osstest service owner wrote:
> > flight 166151 linux-linus real [real]
> > http://logs.test-lab.xenproject.org/osstest/logs/166151/
> > 
> > Regressions :-(
> > 
> > Tests which did not succeed and are blocking,
> > including tests which could not be run:
> >  build-i386-pvops              6 kernel-build             fail REGR. vs. 
> > 165976
> 
> I'm puzzled by this in two ways: First that this is being done at all,
> when 32-bit PV Xen support has gone away several releases back. If the
> purpose is to cover PVH and/or HVM, then I guess the test name has
> become misleading.

"pvops" refers to the branch of Linux, not the test configuration.  It
means "not that weird xenolinux thing".

> And second that this fails on a KVM related build error:
> 
> arch/x86/kvm/x86.c: In function ‘vcpu_enter_guest’:
> arch/x86/kvm/x86.c:9948:1: error: unsupported size for integer register
>  }
>  ^
> make[2]: *** [scripts/Makefile.build:287: arch/x86/kvm/x86.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
> 
> To limit the risk of unrelated build breakage, wouldn't it make sense
> to keep off Kconfig settings which aren't really of interest?

I think "you can't build it with a perfectly normal configuration that
contains both Xen and KVM things" is a relevant thing to know.
Distros need to compile their kernels with many things enabled, some
of which might be mutually exclusive at runtime.

Ian.