Automatically merge with passing tests and a +1 Review

[dstufft]

It would be great if a bot could automatically merge PRs that meet all of the following:

• All status checks are green
• A CLA check is good.
• Someone with write permission to the repository has given it an approved review with no reviews from someone with write permission requesting changes.

Right now if I review someone's PR either one of two things happen:

• The status checks are green and the CLA check is green, in which case I press the merge button.
  • This case is trivial though it does require an extra button click that could be eliminated.
• The status checks are pending or the CLA check hasn't happened yet.
  • This is the case I'm most interested in, what happens here is at some point in the future the status checks will go green but there is no notification that this has happened, so this relies on someone coming around again and notice that the merge button can be pushed.

[ronaldoussoren]

On 15 Feb 2017, at 01:46, Donald Stufft ***@***.***> wrote: It would be great if a bot could automatically merge PRs that meet all of the following: All status checks are green A CLA check is good. Someone with write permission to the repository has given it an approved review with no reviews from someone with write permission requesting changes.

Is having one commiter with an approved review sufficient? There are a number of modules with explicit owners (decimal and itertools spring to mind), for those modules a review of other committers is useful but not necessarily sufficient. Ronald

[brettcannon]

As of right now any one committer is sufficient to clear a PR. If we really wanted to we could have a bot that adds a status check for review approval for modules that have a specific owner (although I personally don't like our personal ownership of modules concept so I won't be putting in the effort to construct such a bot).

[zware]

We could have a convention of "if the touched module is owned by somebody who is particular about it, leave your 'LGTM' comment but don't click the 'approve' button". I don't think a bot for it would be worth the effort.

[zware]

Clarification: I don't think an ownership bot is worth it. I'm all for the auto-merge bot @dstufft proposed.

[dstufft]

@ronaldoussoren If that's the case, it might be a good idea to add a different feature that automatically added those explicit owners as "request a review" for those modules so they are notified there is a PR that touches the module they "own".

I thought at one point I saw a statement that the required reviews thing would mandate a +1 review from anyone that was explicitly marked as "please make a review" but I can't seem to find that now, so it would need tested to see if that would actually cause the PR to be blocked on waiting on a review from them or not.

[dstufft]

I mentioned in #32 about the mention bot that could help notify people about things that need reviewed without having to subscribe to the firehouse, but that same bot could also be used for people who own a particular module to always get notified. It also has a mode that will have the bot do an explicit "request a review" from people, but it looks like that's a global only flag, but we could maybe modify the bot so that module owners can flag themselves to have a review requested of them instead of just being mentioned.

[dhimmel]

Merging methods: squash, rebase, merge

One thing that would be difficult for a bot is how to merge.

For many pull requests, a GitHub squash merge, where all commits are squashed into one with potential manual creation of a consolidated commit message, is the cleanest. For example, python/cpython#97 is a good example of a squashed pull request.

In other cases, it makes sense to use GitHub's merge (or rebase) option. In these cases, the reviewers may direct the pull request contributor to rewrite (and force push) the commit history so it's clean. This is especially important for pull requests with commits from multiple authors. Squashing the entire PR would erase all but one authors' recognition from the commit history. Additionally, there are some larger PRs that are more readable as a series of commits.

[dstufft]

We only allow squash merges on CPython's repository.

[dhimmel]

We only allow squash merges on CPython's repository.

@dstufft that's makes things easier.

In addition, I could imagine a skip-automerge tag or keyword you could add to an issue, that would disable the bot. That would take care of lot's of these difficult & infrequent cases.

[berkerpeksag]

I think time zone differences needs to be considered when a PR is opened by another core developer. If someone approves my PR in 3 am in European time zone and if it breaks some non-Linux buildbots then I wouldn't be able to fix the problem quickly.

[ncoghlan]

@Mariatta pointed out on core-workflow that to do this we'd need some way for a core dev to provide a pre-approved commit message that:

• amends the syntax of the PR reference in the commit title
• provides the actual final commit message, not the concatenation of all the individual commit messages along the way

[Mariatta]

Our newest bot, @miss-islington, can notify the core dev when status checks on a backport PR is completed (she will leave a comment).
So I think there may be ways to get this to work on regular PRs, i.e. leaving a comment that all status checks are complete, so the core dev can come back and merge the PR.
And the comment should only be left when there is awaiting merge label. (meaning a core dev has approved the PR).

[brettcannon]

And based on how the "should core devs merge on other's behalf", probably a label or comment that says "I have signed off on this PR and have no plans to change it later".