[openib-general] [ANNOUNCE] - Chelsio T3 Git Repositories

Steve Wise swise at opengridcomputing.com
Thu Dec 7 14:30:44 PST 2006


On Thu, 2006-12-07 at 14:24 -0800, Sean Hefty wrote:
> > I'm struggling with maintaining a patch series in-review on lklm and
> > netdev, plus maintaining a consistent tree that I can QA on and not
> > introduce bugs from other stuff going into 2.6.20. So I don't want to
> > just base this tree on Roland's for-2.6.20, as an example.  I really
> > just want 2.6.19 + stuff needed to run chelsio's T3.  Right now, that is
> > the UCMA stuff + a few core fixes...
> 
> I'm sure Roland can provide more input here, but what I did was start with 
> 2.6.19.  Then, for each feature set in SVN, I created a new git branch, reworked 
> the SVN patches, and applied them to that branch.  Where I had dependencies, I 
> simply branched off one of my branches.  For example, my multicast branch is off 
>   my rdma_ucm branch.
> 
> My master branch is 2.6.19.  My intent is to update my tree with each new Linux 
> release.
> 
> As an aside, I created a test-apps branch to throw all my kernel test apps into. 
>   (I really didn't want to maintain a branch per test app, since these will 
> never merge upstream.)  I included krping in that tree, since i didn't see where 
> you were maintaining it, and I didn't want to lose it.
> 

Thanks!  I forgot about that stuff!

> > Roland, I welcome your thoughts too on how I should do this.  I'm new to
> > git.  Also I'm using stgit to maintain the chelsio driver patch series,
> > so I continually pop it and add fixes to each patch as I fix things, so
> > the tree really is kind of in-flux...
> 
> I didn't think that you wanted to do this after you've published a tree.  If 
> someone clones your tree, then you use stgit to pop a patch, modify it, then 
> recommit it, I'm not how a cloned tree reconciles the changes.
> 

Well, the life of this T3 git tree will hopefully be short:  we're
trying hard to get the kernel bits of T3 into 2.6.20...

You're right.  Folks cannot back against this tree and do a pull to
refresh.  It'll get balled up.  But Roland's tree is the same way. 


Steve.





More information about the general mailing list