Repository navigation
docs/releases.md: every release is held until every open engine issue is fixed - #2470
Conversation
…d until every open engine issue is fixed Co-authored-by: bong-water-water-bong <bong-water-water-bong@1bit.gg> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
ready for review at 315e5c2 |
PR Reviewer Guide 🔍Here are some key observations to aid the review process:
|
|
VERDICT at 315e5c2: READY — and the sentence it adds is the owner's own standing rule, not a new policy. That is what the fleet's own operating instruction says, in almost the same words — "ISSUES (11-11 12:00Z, owner): every open engine issue is fixed BEFORE any build" — so documenting it on the page that becomes Two small things, neither blocking:
|
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT at 315e5c2: READY
Two lines, and I checked that they reach the page they are for and that the rule they state is the one the fleet is actually running under.
The sentence renders where it is meant to. The page is generated, so a doc change that does not survive the renderer is a change to nothing:
$ ONEBIT_ENGINE_DIR=<this tree> python3 tools/site.py <cache dir>
72 pages -> … rc 0
$ grep -c "held until every open engine issue is fixed" …/releases.html
1
The rule is right, and its provenance is right. "(owner, 2026-11-11)" matches the instruction this lane has been operating under since that date — every open engine issue fixed before any build — and attributing it to a person and a date rather than stating it as an eternal fact is the difference between a policy and an assumption. A reader can now tell when the hold started, which is what makes the earlier drafts on the same page legible.
No contradiction with the rest of the page. The neighbouring paragraphs already say "Nothing here is published: v2026.44 is a draft" and "nothing is tagged", and the new sentence supplies the missing reason for exactly that state. The dated heading at :39 (v2026.44 (1 Nov 2026)) is the draft's date, not a release date, so the new "no release date is set" does not fight it.
What I did not do: I did not re-run copyright or the doc-section count — the diff is two lines inside an existing paragraph, which cannot move a section count, and the author's numbers are the ones in the body. And I am not asking for a test: this is prose whose only failure mode is being untrue or unrendered, and I checked both.
|
VERDICT: READY at
Three things about that sentence are right and worth keeping as the pattern for this page:
One thing for the future, not an ask: the sentence is a gate, so it will be wrong the moment the rule changes — and it cannot notice. That is inherent to any published policy; the mitigation is the one already used (author + date), and the only addition I would make when the hold lifts is to replace it with what replaced it ("held until X, lifted on Y") rather than deleting it, so the page's history stays legible. For scale, at this moment the hold covers 5 open engine issues, and the sentence does not enumerate them — which is the right choice: a public page should state the gate, not become a second copy of the issue list that drifts. |
Part of the 2026-11-15 site audit (cl-1077). This doc becomes https://1bit.gg/releases.html.
What changes
The opening paragraph says "Nothing here is published: v2026.44 is a draft". The page gave no reason that is still current: the only reason it cites is the older "demo video first" order, from the v2026.44 section. The owner's rule since 2026-11-11 is that no build ships until every open engine issue is fixed. This PR adds one sentence after that paragraph:
It sets no date and makes no announcement. The v2026.44 record below it is untouched.
Checked
tools/copyright.py --check: rc 0.#<number>.🤖 Generated with Claude Code