I am also -1 on the bot per the reasons Julian has stated.
Our problem is mainly due to PRs not being reviewed. I think an
extension to the problem is that because PRs take a while to be
reviewed, contributors sometime lose context (and can't find time to
realign themselves with the current state of the codebase) or move on to
another company where the PR is no longer relevant to their new role,
causing PRs to be abandoned.
From what I can see on the jira and Github, most PRs just require a bit
of polish to become mergable. As a contributor, it would be
disheartening to spend time investigating a bug or feature, work on a PR
and have it closed because the project did not have enough resources to
review it (I've had this happen on other opensource projects using stale
bots). In addition, in cases where feedback was given, but the PR was
abandoned by the original contributor, there's often scope for another
contributor to "carry" the PR and the amount of work needed to make it
mergable is not too onerous.
I'd prefer to manually close PRs and JIRAs where the change does not
meet the goals of the project, is of too low quality or have diverged
from master so much (due to age) that creating a PR from scratch is more
manageable.
On 1/03/2019 6:19 am, Julian Hyde wrote:
-1 Using a robot to close stale PRs is solving the wrong problem.
The main reason that we have a lot of open PRs is that we - as a community of
committers - are not putting sufficient time into reviewing. I think that we
are giving PR submitters poor service, and they are being very patient with us.
Every PR is a potential new community member, and not one but many improvements
to Calcite. Our goal, as committers, should be to ensure that we don’t let any
of those opportunities slip.
We all know that there are a few poor quality PRs. They take more effort to
review, and bang into shape, than it would have taken the reviewer to write it
themselves. Comments are misspelled, and the code just looks sloppy. Those,
frankly, are not very high priority.
But at the other end of the spectrum, there are some - many - high quality PRs,
where the contributor has done their homework. They are rarely perfect on the
first submission, but you can see that the person behind it would make the
community much stronger. Those PRs deserve our immediate attention.
And I’ll say this just once. I get frustrated when committers prioritize PRs
from employees of their own company, and I get furious when they hold those PRs
to a lower standard than other PRs. As a committer, you need to wear your
Apache hat, and only your Apache hat, when reviewing code.
Julian
On Feb 28, 2019, at 6:28 AM, Michael Mior <[email protected]> wrote:
Thanks Kevin! That's an important task and one that few people are
willing to take on :)
And thanks Hongze for the pointer to Stale. Especially since other
Apache project are already using it, I'd be inclined to have a
discussion on the appropriate configuration and give it a go.
Personally, I think the list of open PRs should things that are
actively being worked on. Closed PRs can always be reopened anyway, so
I don't think we're losing anything.
--
Michael Mior
[email protected]
Le mer. 27 févr. 2019 à 14:36, Kevin Risden <[email protected]> a écrit :
There are 105 open pull requests against apache/calcite repo [1]. There are
only 48 Calcite JIRAs labeled with pull-request-available [2].
I'm planning to go through in the next few days and make sure that we have
PRs that match open JIRAs and are labeled pull-request-available. If there
are PRs that are open for JIRAs that are closed, planning to close those
PRs with a comment.
[1] https://github.com/apache/calcite/pulls
[2]
https://issues.apache.org/jira/issues/?jql=project%20%3D%20CALCITE%20AND%20resolution%20%3D%20Unresolved%20AND%20labels%20%3D%20pull-request-available%20ORDER%20BY%20priority%20DESC
Kevin Risden
