Repository navigation
Conversation
4b8ed1c to
c34e788
Compare
|
|
a681b72 to
8af40b8
Compare
96ffdb2 to
3a6799e
Compare
3a6799e to
65e2f5a
Compare
65e2f5a to
36ac2dd
Compare
|
On the community meeting, it was mentioned that the top description is actually no longer up to date:
However, personally I think it is simpler to just rely on github's auto-closing of issues based on a "closes #.." (or equivalent). IMO it is simpler (for the maintainer, because that is in general how github works, and you don't have to know all details of our custom workflows) and avoids having additional custom workflow logic (it also allows the maintainer for whatever reason to not close the issue by merging the PR, or if we in the future want to use this practice more) |
|
And while we are touching the merge workflow, is there appetite in discussing what the commit message should be?
Would we be fine with just using the title (any co-authored-by lines as needed)? |
dev/release/post-04-website.sh published a partial count if one of the GitHub requests failed. Also document that the script needs an authenticated gh.
Address review comments: rely on GitHub closing the issue from a closing keyword in the pull request description. * The Boring Cyborg app writes "Closes #XXX" to the "GitHub Issue:" line of a pull request to main. It leaves "Refs #XXX" alone: that is for a partial fix, which keeps the issue open. * The "Dev PR merged" workflow doesn't close the issue or comment on it anymore. It only assigns the issue and sets its milestone, and only if GitHub closed the issue. It doesn't need the "(#YYY)" suffix of the commit title anymore.
Changed to use "closes #" 5619a5b. I like it's the github way of doing thing, but not sure either way. |
Why would we not store a long message and keep git as the canonical history record (even if sometimes wildly wrong). It's possible to reduce verbosity of |
raulcd
left a comment
There was a problem hiding this comment.
We could add a warning on the merge PR script suggesting committers to use the merge button now.
Co-authored-by: Raúl Cumplido <raulcd@apache.org>
e4c5fb7 to
fe4116a
Compare
| # because of "* GitHub Issue: Closes #XXX" in the pull request | ||
| # description. This sets what GitHub doesn't set: the assignee | ||
| # and the milestone of the issue. | ||
| - name: Update issue |
There was a problem hiding this comment.
Maybe a minor nit for easier debugging could be to split this Update issue into ~3 steps:
- Get PR and issue
- Assign issue
- Assign milestone
So it's easier to see where it failed / debug steps etc.
There was a problem hiding this comment.
Split into three steps in 20eb0e7: "Get issue", "Assign issue" and "Set milestone". The first step is "Get issue" and not "Get PR and issue" because the workflow reads the issue number from the commit title and does not fetch the pull request. The conditions are now if: of the steps, so a step that has nothing to do is shown as skipped.
Rationale for this change
Committers need a personal access token and
dev/merge_arrow_pr.pyto merge a pull request. With this change they can use the "Squash and merge" button instead.What changes are included in this PR?
.github/boring-cyborg.yml: for a pull request tomain, the Boring Cyborg app writesCloses #XXXto theGitHub Issue:line of the description, so GitHub closes the issue on merge. It leavesRefs #XXXon that line alone: that keeps the issue open for a partial fix. The "PR Title" status no longer acceptsWIP: GH-XXXtitles. It doesn't block the merge..github/workflows/dev_pr_merged.yml(new): does the rest of the issue update of the merge script. On a push tomainwith aGH-XXX: ...commit title, if GitHub closed the issue, it assigns the issue to the commit author if nobody is assigned and sets the next major release as the milestone if the issue has none.dev/release/post-04-website.sh: "Patch Committers" is counted from who merged each pull request on GitHub. Button merges don't have theSigned-off-by:trailers it used.Are these changes tested?
CI passes, but it doesn't run the new workflow: that can only run once this is on
main. Locally, I ran the workflow step against apache/arrow with the writes printed instead of executed, checked the new patterns of the Boring Cyborg configuration against sample titles and descriptions, and compared the new committer count for 25.0.0..26.0.0 with the old one (395 merges, 390 from the trailers).Are there any user-facing changes?
No.
Was AI used for this PR?
PR code and description written by:
Reviewed before submission by: