Re: [Xen-devel] [PATCH 3 of 5] mem_access: public interface changes

[Joe Epstein <jepstein98@xxxxxxxxx>]

Sorry about that.  I had considered the whole series as one logical change, and broke it up by sets of directories to reduce the patch size and yet kept the changes local so that there aren't too many places to look when verifying that the patch went in correctly.  I admit that that ordering isn't the easiest to read through.


I'll refactor the chunks for both this series and the mem_access 2 series into more paired up units and resubmit them today.  It will be identical code, for anyone who has applied the ones already submitted, but will be in intermediately compilable patches.  The patches will end up being longer for that, so I hope that's okay.  The other alternative is one long patch, which I will attach in the alternative to my 0 of X email to make it easier to apply.


Thanks

On Tue, Jan 4, 2011 at 6:45 AM, Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote:

I didn't review the whole series I just happened to notice this while
reviewing the use of hypercall buffers in libxc:

 On Wed, 2010-12-29 at 07:27 +0000, Joe Epstein wrote:
> +struct xen_hvm_set_mem_access {

The definition of this structure needs to come along with the first use
of it, which in this case would be the 1/5 patch which adds the
hypercall.

Similarly for any other interface changes etc. You need the code changes
to be with the interface changes, or otherwise to construct your patch
series such that you add the new interface, convert everything to it and
then remove the old interface.

Otherwise everything is broken mid-way through the series, it should
compile and work at each step/patch.

It also makes review hard since each patch doesn't really stand alone.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel