On Tue, Jul 12, 2016 at 01:15:44AM -0400, Boris Zbarsky wrote: > On 7/12/16 1:02 AM, Mike Hommey wrote: > > The envisioned workflow is that you don't rebase on inbound to push, > > since autoland handles the landing for you. So you can keep your tree as > > it was before you wanted to land. > > OK, what about the case when A was not authored by me?
If you have B and C depending on A that was not authored by you, it's probably already in your local branch, isn't it? If you're waiting for A to land before working on B and C, the case is not very different whether A lands on inbound or on central through autoland: you have to wait for someone to do something ; in one case, you wait for someone to land A on inbound, in the other, you wait for a sheriff to merge autoland. You wait different persons, but you still wait for someone. > What's the expected frequency of autoland merges? I don't have this information. Eventually, when integration branches disappear, I guess what happens is that autoland gets merged to central as soon as test results are all up for the push. Or if there's enough confidence that there's no new intermittent because of it. That's actually an interesting question. We want to avoid backouts, but there's going to have to be a balance between having the landings on central quickly enough and having confidence they don't break the tree. And at some point some subtle thing will break but will have made it to central. Mike _______________________________________________ dev-builds mailing list dev-builds@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-builds
