[ovs-dev] Removing "xs5.7" Branch?
Ian.Campbell at citrix.com
Mon Apr 12 09:17:51 UTC 2010
On Mon, 2010-04-12 at 08:55 +0100, Justin Pettit wrote:
> Hey, Ian and Ben. As you've probably noticed, I sent out some patches
> for things that were never transitioned from the "xs5.7" branch. I've
> done my best to make it so that these changes work with XenServer 5.5
> and 5.6, since OVS needs to support both.
> The only piece that appears to be missing is "fcc4950: Improve
> xsconsole plugin in various ways." Everything from this commit should
> be brought over except the "xapi now takes care of syncing the setting
> to all hosts so no need to do that manually" part. Is this only true
> for 5.6, since "vswitch_controller" is a first class datamodel field?
> For 5.5 should we still manually do the sync, since we use the
> "other_config" pool key? If so, I guess I'll just leave the code and
> not execute it in 5.6. Does this seem reasonable?
That sounds like the right approach to me. Looks like you can just make
the calls to Task.Sync + self._update* conditional?
> In the bring_down() function in InterfaceReconfigureVswitch.py,
> there's some commented out code regarding deconfiguring network-uuids.
> Do we still need to actively do that? We should either fix it or
> remove the code.
I'm not sure what is going on here. It looks like you uncommented it in
2a577bd8 (with my ack) and I commented it back out (probably by
accident) in b3080599.
interface-reconfigure should just be doing what is needed by the
vswitch/controller here so I'm happy for you to decide what that is.
> Once we get these things resolved, do you feel comfortable scrapping
> the "xs5.7" branch?
I am comfortable. I never actually used the branch myself (since it was
never really up to date with master or next) and just cherry-picked the
changes I needed into my local patch queue.
> I'd appreciate you take a quick look to make sure I didn't miss
I took a quick look at the output of
$ git log master..origin/xs5.7
(which I think shows commits in origin/xs5.7 but not in master, the
actual output supports this theory) and didn't see anything which wasn't
either already in master or covered by the series you posted this
Is the citrix branch also a dead-end?
More information about the dev