xen changeset is not available

[Zhao, Yunfeng]

Xen changeset is not available on FC6 Test1.
And this info is very useful for fedora-xen user.

Is it possible to build changeset number in xen binary of FC6?

Thanks
Yunfeng

[Gawain Lynch]

On Wed, 2006-07-05 at 10:13 +0800, Zhao, Yunfeng wrote:
>  Xen changeset is not available on FC6 Test1.
>  And this info is very useful for fedora-xen user.
>  
>  Is it possible to build changeset number in xen binary of FC6? 
I have to agree with this one.  It would be useful to me too.

[Jeremy Katz]

On Wed, 2006-07-05 at 10:13 +0800, Zhao, Yunfeng wrote:
>  Xen changeset is not available on FC6 Test1.
>  And this info is very useful for fedora-xen user.
>  
>  Is it possible to build changeset number in xen binary of FC6? 
There isn't just one changeset, though.  The HV is one changeset of
-unstable, the kernel patch comes from the hg tree that tracks Linus's
tree (and thus doesn't have a directly applicable changeset) and the
tools are a third.  Although hopefully for releases, the tools are
actually a _released_ version of the tools rather than random changeset
of the day.

So the single "one size fits all" changeset mentality of Xen doesn't
really work.  We do try to note the changeset in the spec
files/changelogs when doing updates.

Jeremy

[Aron Griffis]

Jeremy Katz wrote:  [Wed Jul 05 2006, 10:01:16AM EDT]
>  On Wed, 2006-07-05 at 10:13 +0800, Zhao, Yunfeng wrote:
>  > Xen changeset is not available on FC6 Test1.
>  > And this info is very useful for fedora-xen user.
>  > 
>  > Is it possible to build changeset number in xen binary of FC6?
>  
>  There isn't just one changeset, though.  The HV is one changeset of
>  -unstable, the kernel patch comes from the hg tree that tracks
>  Linus's tree (and thus doesn't have a directly applicable changeset)
>  and the tools are a third.  Although hopefully for releases, the
>  tools are actually a _released_ version of the tools rather than
>  random changeset of the day. 
IMHO this is the wrong approach.  It's already bitten ia64, and it's
bound to bite other architectures at some point.  See my comments at
https://www.redhat.com/archives/fedora-xen/2006-June/msg00154.html
along with Chris's response.  Fedora should be using matched versions
of hypervisor, xenlinux patch and tools, at least as much as possible.
In any case, providing the xen-unstable changeset on which each is
based would be helpful to all parties involved.

Granted, the linux-2.6-xen.patch comes from linux-2.6.tip-xen rather
than xen-unstable directly, but if you trace it back, you can find out
the xen-unstable changeset on which it's based.  At that point,
snapshotting the hypervisor to match becomes relatively easy, and so
does providing the changeset in the release/changelog/description/etc.

Here's an example, if it's helpful:

    hg clone http://xenbits.xensource.com/ext/linux-2.6.tip-xen.hg
    cd linux-2.6.tip-xen.hg
    xen_unstable_cset=$(hg log -v | awk -F '[ :]' \
        '/^xen-unstable changeset:/{print $(NF-1);exit}')

So then, make a matched hypervisor:

    hg clone http://xenbits.xensource.com/ext/xen-unstable.hg
    cd xen-unstable.hg
    hg up -C $xen_unstable_cset
    tar cjf ../xen-$xen_unstable_cset.tar.bz2 xen

Regards,
Aron