And another way suggested by Jon’s statement, is to not squash the commit, but
leave at least one of each committer’s contributions intact. That might not be
too hard to add to the prepare-commit script, it’s just an invocation of
interactive rebase/squash.
On 7/5/17, 10:32 AM, "Matt Foley" <mfo...@hortonworks.com> wrote:
Jon, good point. I was thinking we would just use the commit message to
share attribution.
It would be nicer to do something that would be picked up by github’s
commit attribution, but unfortunately there is only one “author” field in a git
commit metadata. There are discussions online about allowing multiple
instances of that field, and the maintainers don’t seem willing to do so
currently. There seems to be considerable support in the community for putting
a LIST of authors in that single field; see:
https://stackoverflow.com/questions/7442112/attributing-a-single-commit-to-multiple-developers
and perhaps even this tool:
https://github.com/chrisk/git-pair
but it isn’t clear that github can correctly extract the email addresses in
that case. Github uses the metadata emails to attribute authorship, see
https://help.github.com/articles/updating-commit-author-attribution-with-github-importer/
I’d be willing to experiment a little to see if github can pick up multiple
emails, but I wouldn’t bet on it.
Another way to do it would be with trivial commits following the real
commit, one per additional author besides the committer. This method would
count commits correctly, but not LOC.
--Matt
On 7/5/17, 6:22 AM, "zeo...@gmail.com" <zeo...@gmail.com> wrote:
That all sounds pretty reasonable to me. My biggest concern would be
attribution during step 5 - we would need to make sure it isn't squash
merged like we typically do (assuming we do properly squash merge into
the speculative
branch). Not a big issue though, I guess, just need to make sure it
isn't
overlooked.
Jon
On Wed, Jul 5, 2017 at 4:40 AM Matt Foley <ma...@apache.org> wrote:
> Now that METRON-877 is in, I would like to proceed with Steps 3-6 of
the
> remaining work to separate out Stellar functionality as an independent
> module. A couple people have suggested that this further development
> should be done in a Metron “dev branch”, where:
> a) changes are more visible than in a single person’s private
development
> branch, and
> b) work can proceed for several days or a couple weeks on a branch
that
> the collaborators may choose to keep stable for the duration (ie,
without
> constantly updating to master).
>
> This concept was discussed as a “speculative branch” in this email
thread:
>
https://lists.apache.org/thread.html/391e15347ad625c4aa61e81f5dd238c0acb4048b8d77f93313298263@%3Cdev.metron.apache.org%3E
> but I don’t see that we ever actually changed our bylaws to mention
it.
>
> Nevertheless, it falls within the purview of the PMC to create new
> branches in our code tree, and I request PMC members to give me a lazy
> consensus vote to do so. Please +1 this email if you agree.
>
> The proposed rules of engagement are (drawn from issues raised in that
> email thread):
> 1. Commits to this branch to have the same rules as to master: Jira,
PR,
> and at least one +1 from a knowledgeable reviewer, and no -1’s.
> 2. +1 reviews may come from any participating contributor, not only
> current committers. But commits still have to be made by a
committer, so
> we don’t have to create new auth infra for this branch.
> 3. The branch should be updated from master at least every second
week, or
> more frequently. This may be adjusted to avoid disruption of work in
> progress.
> 4. PR’s to master will be posted for review as soon as self-consistent
> chunks of useful functionality are done. The collaborators will
define
> those chunks, but a rough goal is every two weeks. The goal is to
avoid
> mega-patches to review.
> 5. PR’s to master will be posted by a single developer from their home
> github repo, not directly from the speculative branch, so that
> collaborative work can proceed on the speculative branch.
> 6. The PR’s will be credited equally to all collaborators active
during
> that “chunk” of work.
> 7. PR’s to master have to be reviewed and agreed to as though they
were
> new patches. (The fact they were previously accepted into the
speculative
> branch is at most a recommendation, not an a priori decision to let
them
> into master.) The usual rules apply. While collaborators will likely
want
> to +1 such PRs, sufficient time must be provided for other community
> members to review and raise issues.
>
> Thanks,
> --Matt
>
>
>
> --
Jon
