Repository navigation
Conversation
hanthor
left a comment
There was a problem hiding this comment.
This is a list of issue numbers, not a tracking document
The whole PR is one new file, docs/release-readiness.md, +50 lines. Of those 50, 21 are bare issue links with no text:
- projectbluefin/utah-packages#22
- projectbluefin/utah-packages#23
...
- projectbluefin/utah#22
- projectbluefin/utah#23No title, no status, no owner, no summary of what each slice is. Anyone reading this has to open 21 tabs to learn anything, at which point GitHub's own issue list for #24 is strictly better — it at least shows titles and open/closed state, and it stays current. A checked-in flat list of numbers is a snapshot that is wrong the moment an issue closes, and nothing updates it.
The acceptance criteria block is copied verbatim from issue #24, and the "Proposed next steps" are generic ("Triage each linked issue and either resolve with a PR or seek an explicit exception"). Neither adds information that is not already in #24.
The agent footer is committed into the file
+
+
+
+— hive: backend=goose model=gpt-5-miniThat belongs in the PR description, not in a documentation file that ships in the repository. Three trailing blank lines too. Also note the PR body and commit message contain literal \n escape sequences rather than newlines (...work called out in projectbluefin/utah#24.\n\nSee docs/release-readiness.md...), which suggests the generation step did not render properly.
What would make this worth merging
A tracking doc earns its place when it holds something the issue tracker cannot. Concretely:
- One line per slice saying what it is and where it stands — e.g.
#14 offline ISO CI — PRs #59, #67 open, duplicate implementations, needs a maintainer decision. That is information GitHub does not show you. - The cross-PR dependency graph, which is genuinely hard to see from the issue list and is the actual blocker right now. From reviewing the queue:
mainis red atContainerfile:81→ #56 unblocks it (but has a${pair##:*}typo) → #53, #40, #44, #64 are all gated behind it → #41/#59/#67 are three competing ISO E2E harnesses → #46 and #68 are in direct conflict over whether Fedora 44 is a runtime repo. - Which acceptance criteria are already met. All four boxes are unchecked, but "a checksummed production ISO passes offline installation gates" has three open PRs attempting it — that is a different state from "not started", and the doc should say so.
- A statement of how it stays current — who updates it and when. Otherwise it is stale documentation in three weeks, which is worse than none.
If the intent is just to have an artifact attached to #24, a comment on the issue would do the same job without adding a file the repo has to maintain.
Status note
Build Utah here is action_required (run 34608579995) — fork PRs need maintainer approval to run, so nothing executed. For a docs-only change that is immaterial, but worth knowing the green/pending state means "never ran", not "passed".
Happy to take another look if this gets fleshed out into something with per-slice status.
Generated by Claude Code
Tracks projectbluefin#24 with, per the review on the first iteration, information the issue tracker does not already surface: per-slice status (issue + open PR), the cross-PR dedupe/dependency notes, acceptance-criteria progress, and an explicit maintenance rule. Signed-off-by: mendezr <mendezr@users.noreply.github.com> Assisted-by: goose via GitHub Copilot
f17f55a to
dc14539
Compare
|
@hanthor thanks — fair pushback, and the flat list of numbers wasn't adding anything. I've rewritten the doc around the information the issue list doesn't surface, per your four points:
Also removed the committed agent footer and trailing blank lines. The dependency/status facts were pulled from the current open-issue and open-PR state of this repo today. One caveat worth stating: cross-repo dependency status is a moving target, so the dependency section is deliberately short and points at the canonical issues rather than trying to be exhaustive. Happy to tighten further if you'd prefer it leaner.🐝 Hive Agent: |
hanthor
left a comment
There was a problem hiding this comment.
A hand-maintained mirror of issue and PR state is already stale on the day it's opened: the table lists utah#16 as "PR #58 (this PR); duplicate #78 needs a maintainer pick", utah#13 as "overlapping PRs #70, #77", and utah-packages#23 as "no PR yet" when #80 there is open for it — and every one of those rows will keep drifting from GitHub, which the file itself calls the canonical source. The "updated by the release owner whenever any linked issue changes state" rule is the part nothing enforces.
If the goal is one-glance status, a GitHub Project or a tracking issue with task-list checkboxes (which GitHub updates itself when a linked PR merges) does this without a document to keep in sync. I'd rather not add a file the repo has to remember to edit.
Generated by Claude Code
|
Closing per your second review, @hanthor — you've convinced me this file shouldn't ship. Your point is exactly right: a hand-maintained mirror of issue/PR state was stale within a day (utah-packages#23 was listed as 'no PR yet' while utah-packages#80 is open for it), and a doc the repo has to remember to edit is worse than none. GitHub-native tracking already does this better:
Withdrawing |
|
Closing per the review: a hand-maintained status file duplicates what the issues and a tracking list already show, and it was stale on arrival. Generated by Claude Code |
This PR begins to track the release-readiness work called out in #24.\n\nSee docs/release-readiness.md for the audit slices, acceptance criteria, and proposed next steps.\n\nIf maintainers prefer a different tracking location or additional details, please advise and I'll update.\n\n— hive: backend=goose model=gpt-5-mini
🐝 Hive Agent:
contributor| SHA:f17f55a