From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alex Williamson Subject: Re: [GIT PULL] VFIO updates for v3.8 Date: Tue, 11 Dec 2012 19:59:32 -0700 Message-ID: <1355281172.3224.188.camel@bling.home> References: <1355252887.3224.176.camel@bling.home> <20121212104632.b01520d6015486dd3ae507ef@canb.auug.org.au> <1355270816.3224.179.camel@bling.home> <20121212114430.cceef241f595a95e16c48874@canb.auug.org.au> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: Linus Torvalds , linux-kernel , kvm , fengguang.wu@intel.com, julia.lawall@lip6.fr, jiang.liu@huawei.com To: Stephen Rothwell Return-path: In-Reply-To: <20121212114430.cceef241f595a95e16c48874@canb.auug.org.au> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org Hi Stephen, On Wed, 2012-12-12 at 11:44 +1100, Stephen Rothwell wrote: > Hi Alex, > > On Tue, 11 Dec 2012 17:06:56 -0700 Alex Williamson wrote: > > > > Is that a bad thing? I can start tagging from my next branch if that's > > preferred. Thanks, > > Linus has said many times to not rebase before sending a pull request. > When you rebase your tree you effectively throw away your testing (since > the thing you rebased on top of may have introduced semantic conflicts > with the work in your tree). If you don't rebase your tested tree, any > conflicts are then restricted to the actual merge and can be fixed there > (or at least the diagnosis will lead there). > > So, if I was a maintiner, at the start of the merge window (or just > before) I would create a test branch that contained my work plus a > *merge* with Linus' tree and do some testing on that and then ask Linus > to pull my tree (not the merged version). It may prove that the test > merge with Linus' tree produces an "interesting" syntactic conflict - in > this case I would mention that to Linus and put the merged tree somewhere > public for him to use as a guide. (Mind you, this conflict would already > have most likely been noted by the linux-next maintainer.) > > Also, your testing may have brought to light a semantic conflict, in > which case the fix could be supplied to Linus with the pull request, or a > well changed logged back merge of Linus' tree containing the fix could be > done and Linus asked to pull the result. Thanks for the tip. I certainly retested after doing the rebase to v3.7, but I can see the point. I'll do as you suggest, a merge on a separate branch for testing only and tag what I currently have in my next branch. v2 forthcoming. Thanks, Alex