Re: [Xen-devel] [linux-linus test] 25753: regressions - trouble: blocked/broken/fail/pass

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

David Vrabel writes ("Re: [Xen-devel] [linux-linus test] 25753: regressions - 
trouble: blocked/broken/fail/pass"):
> On 05/04/14 02:41, xen.org wrote:
...
> >  test-amd64-i386-xl            5 xen-boot                  fail REGR. vs. 
> > 12557

> Linux is currently broken for 32-bit x86 PV guests due to some changes
> to the kernel/irq stacks.  I think Boris is investigating.

In the above test name we have, firstly, the hypervisor architecture,
and then the dom0 and guest architecture.  But the test names are
condensed and don't necessarily encode everything relevant:

> I don't know if this explains these failures.  It's not obvious what the
> test name translates to.

The "meaning" of the test name is really a set of variables which
control the test run, which can be seen here (e.g.):
   
http://www.chiark.greenend.org.uk/~xensrcts/logs/25753/test-amd64-i386-xl/info.html
(linked to from the column heading for the test in question in the
matrix in the results mail you were replying to).

The key variables which demonstrate that the bug you describe would
apply to this test are these:

  arch = i386    dom0 architecture, ie what Xen calls x86_32

  buildjob = build-i386
            specifies implicitly the tools architecture which necessarily
            is the same as the dom0 architecture [1]

  xenbuildjob = build-amd64
            specifies implicitly the hypervisor architecture ie
            what Xen calls x86_64

  debian_arch = i386
            architecture for the guest named "debian"

[1] In principle it might be necessary to look at
   http://www.chiark.greenend.org.uk/~xensrcts/logs/25753/build-i386/info.html
and observe "arch=i386" to see that this is a 32-bit build, but
the job name's implications are pretty straightforward.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel