On Fri, Sep 18, 2015 at 8:47 AM, Chen Li <[email protected]> wrote: > I assume the conclusion is: we keep our current git practice. Right? >
I didn't mean to assert this. I just wanted to share the history and the Gerrit limitations. There's no question that things are the way they are today due in part to my personal prejudices, but mine is definitely not the only voice that matters. > Based on this practice, is there any easy way for people like Jianfeng > to make their merge into their branch simpler? I'm afraid I don't think there is. If you are working on multiple feature branches simultaneously, that shouldn't be directly impacted by this. But if you are trying to maintain several *dependent* branches locally (ie, you create branch "foo" from your local master, and then create branch "bar" from "foo") and then merge them independently to Gerrit, things are going to go badly sideways for you. The only way I can think that this working method would be functional would be if we merged all feature branches to actual branches on Gerrit, and used only merge commits when merging onto master. To be honest I'm not sure I could easily enumerate the pros and cons of this approach. > I think Young-Seok is > doing experiencing something similar to merge the master into his > one-year-behind geo branch. For what it's worth, I don't think any different merging strategy would have made much difference in that case. When you fall that far behind master, it's very likely going to be a big job to catch up. Ceej aka Chris Hillery
