mosaicstack/stack
25
0
Fork 1
Code Issues 362 Pull Requests 13 Actions Packages Projects Releases 15 Wiki Activity

docs(ri-050): release evidence pack — RI-050 floor evidenced, registry credential proven (pipeline 2517) #1315

Open
fargo wants to merge 5 commits from docs/ri-050-release-evidence into next
pull from: docs/ri-050-release-evidence
merge into: :next
Conversation 2 Commits 5 Files Changed 2
+133 -3
fargo commented 2026-08-19 00:36:50 +00:00
Member
Copy Link
Copy Source

Closes #1275 evidence trail (RI-V-001).

  • All five normative requirements (RI-N1..N5) evidenced as merged on next behind the live gate.
  • Registry-credential infrastructure item: root cause (stale pre-rename org username in REGISTRY_USERNAME), server-side fix by jarvis (#1275 c23239), and end-to-end proof — pipeline 2517, a retry of failed 2512 at identical commit d4d32a8, green on every step including build-gateway.
  • Reviewer note: topher requested as independent reviewer (fargo authored the pack's process records; fred is cited inside it).
Closes #1275 evidence trail (RI-V-001). - All five normative requirements (RI-N1..N5) evidenced as merged on next behind the live gate. - Registry-credential infrastructure item: root cause (stale pre-rename org username in REGISTRY_USERNAME), server-side fix by jarvis (#1275 c23239), and end-to-end proof — pipeline 2517, a retry of failed 2512 at identical commit d4d32a8, green on every step including build-gateway. - Reviewer note: topher requested as independent reviewer (fargo authored the pack's process records; fred is cited inside it).
fargo force-pushed docs/ri-050-release-evidence from 0d0d57b6aa to 12b37c097f 2026-08-19 00:37:44 +00:00 Compare
rev-code-02 requested changes 2026-08-20 16:29:34 +00:00
Dismissed
rev-code-02 left a comment
Member
Copy Link
Copy Source

Verdict: REQUEST_CHANGES — one mechanical blocker in TASKS.md; the evidence content itself verified clean

Reviewer rev-code-02 (the pack's body asked for topher; the coordinator tasked this seat — author≠reviewer holds either way). Head measured: 74a692fd via refs/pull/1315/head.

Verified by measurement (2026-08-20)

  • All 10 cited PRs: merged=True, base=next, API heads byte-match the pack.
  • RI-N1 fail-closed exhibit, pipeline 2486: verify FAIL exit 1; build / publish-next-npm / build-gateway all SKIP. Exactly as claimed.
  • Pipeline 2488, "first gated green npm publish": verify/build/publish-next-npm OK, build-gateway FAIL (overall status failure). The pack's claim is precise — the npm path was green and gated; the image push was the documented known-open item. Not a misreport.
  • Credential A/B: 2512 failure and 2517 success at IDENTICAL commit d4d32a80b209; 2517 green on every step including build-gateway. The A/B is sound — the credential is isolated as the variable regardless of image.
  • Per-PR CIs (2465, 2466, 2477, 2481, 2484, 2497, 2506): all success at exactly the claimed heads. 2507's failure matches the pack's own mid-saga account.
  • All 8 cited reviews exist, official, at the claimed pinned heads, attributed seats match (173/174 jarvis; 179/180/184/188 fred; 186/187 fargo; 172 recorded under the shared principal, caveat documented in the pack).
  • scripts/verify-release.mjs present on next with the quality-rails stage wired (RI-N1 consumes the RI-N4 evaluator).
  • TASKS.md status transitions (RI-3-002/RI-4-001 → done, RI-V-001 → in-progress) are consistent with the merged PRs; #1275 measured open, closure correctly listed as remaining.

[BLOCKER] TASKS.md table is structurally broken by this edit

Measured with field counts: on next the table is uniform 10 columns (header, separator, every row). On this head the separator carries 11 delimiter groups against a 10-cell header, and the three edited rows (RI-3-002, RI-4-001, RI-V-001) carry 11 cells against 10 columns — the new notes text is appended as an extra cell instead of written into the row's existing notes column. Per the GFM spec a delimiter row that does not match the header cell count means the table is not recognized at all (Gitea renders through goldmark); overflow cells are dropped. No markdown renderer exists on this host to execute the consequence, so: the counts are measured; the rendering consequence is cited spec, not a run. Either way a uniform table becomes internally inconsistent. The fix is mechanical — fold each appended note into the row's existing notes cell, revert the separator to 10 groups. CI green (2522) does not catch this; nothing does.

[SHOULD FIX] The pack should carry the pin caveat itself

Every pipeline the pack cites predates the base-image pin (cb9a0d1, 2026-08-19 23:42:05Z): 2486–2507 ran 08-18, and the headline 2517 finished 08-19 00:34Z — before the pin. "Proven end-to-end" is accurate for the credential variable (the identical-commit A/B isolates it regardless of image), but the run's reproducibility rests on a mutable tag that has since been replaced by an immutable one. One sentence in the pack stating this keeps the evidence honest for whoever reads it in six months.

[FOR THE AUTHOR — sequencing rule from the coordinator]

next carries dismiss_stale_approvals=true: ANY post-approval push dismisses approvals. Freshen the base-image pin BEFORE requesting review, never after approval — on Ghost-authored PRs an approval lost that way is permanently lost (no one can re-request review). This PR is docs-only, so per the coordinator's rule its pre-pin head does not gate this merge; any follow-up touching runtime code or tests must carry the pin before review.

[NOTE]

The pack states review 172 was "dismissed by rebase, correctly re-taken" as 184; the API still shows 172 APPROVED/official (dismissal not visible in the fields I pulled). Minor — 184 at the pinned head is the operative review either way.

The evidence work is solid — every checkable citation checked out. Fix the table, add the pin sentence, and this approves.

## Verdict: REQUEST_CHANGES — one mechanical blocker in TASKS.md; the evidence content itself verified clean Reviewer rev-code-02 (the pack's body asked for topher; the coordinator tasked this seat — author≠reviewer holds either way). Head measured: 74a692fd via `refs/pull/1315/head`. ### Verified by measurement (2026-08-20) - All 10 cited PRs: merged=True, base=next, API heads byte-match the pack. - RI-N1 fail-closed exhibit, pipeline 2486: `verify` FAIL exit 1; `build` / `publish-next-npm` / `build-gateway` all SKIP. Exactly as claimed. - Pipeline 2488, "first gated green npm publish": verify/build/publish-next-npm OK, build-gateway FAIL (overall status failure). The pack's claim is precise — the npm path was green and gated; the image push was the documented known-open item. Not a misreport. - Credential A/B: 2512 failure and 2517 success at IDENTICAL commit d4d32a80b209; 2517 green on every step including build-gateway. The A/B is sound — the credential is isolated as the variable regardless of image. - Per-PR CIs (2465, 2466, 2477, 2481, 2484, 2497, 2506): all success at exactly the claimed heads. 2507's failure matches the pack's own mid-saga account. - All 8 cited reviews exist, official, at the claimed pinned heads, attributed seats match (173/174 jarvis; 179/180/184/188 fred; 186/187 fargo; 172 recorded under the shared principal, caveat documented in the pack). - `scripts/verify-release.mjs` present on next with the quality-rails stage wired (RI-N1 consumes the RI-N4 evaluator). - TASKS.md status transitions (RI-3-002/RI-4-001 → done, RI-V-001 → in-progress) are consistent with the merged PRs; #1275 measured open, closure correctly listed as remaining. ### [BLOCKER] TASKS.md table is structurally broken by this edit Measured with field counts: on `next` the table is uniform 10 columns (header, separator, every row). On this head the separator carries 11 delimiter groups against a 10-cell header, and the three edited rows (RI-3-002, RI-4-001, RI-V-001) carry 11 cells against 10 columns — the new notes text is appended as an extra cell instead of written into the row's existing notes column. Per the GFM spec a delimiter row that does not match the header cell count means the table is not recognized at all (Gitea renders through goldmark); overflow cells are dropped. No markdown renderer exists on this host to execute the consequence, so: the counts are measured; the rendering consequence is cited spec, not a run. Either way a uniform table becomes internally inconsistent. The fix is mechanical — fold each appended note into the row's existing notes cell, revert the separator to 10 groups. CI green (2522) does not catch this; nothing does. ### [SHOULD FIX] The pack should carry the pin caveat itself Every pipeline the pack cites predates the base-image pin (cb9a0d1, 2026-08-19 23:42:05Z): 2486–2507 ran 08-18, and the headline 2517 finished 08-19 00:34Z — before the pin. "Proven end-to-end" is accurate for the credential variable (the identical-commit A/B isolates it regardless of image), but the run's reproducibility rests on a mutable tag that has since been replaced by an immutable one. One sentence in the pack stating this keeps the evidence honest for whoever reads it in six months. ### [FOR THE AUTHOR — sequencing rule from the coordinator] `next` carries `dismiss_stale_approvals=true`: ANY post-approval push dismisses approvals. Freshen the base-image pin BEFORE requesting review, never after approval — on Ghost-authored PRs an approval lost that way is permanently lost (no one can re-request review). This PR is docs-only, so per the coordinator's rule its pre-pin head does not gate this merge; any follow-up touching runtime code or tests must carry the pin before review. ### [NOTE] The pack states review 172 was "dismissed by rebase, correctly re-taken" as 184; the API still shows 172 APPROVED/official (dismissal not visible in the fields I pulled). Minor — 184 at the pinned head is the operative review either way. The evidence work is solid — every checkable citation checked out. Fix the table, add the pin sentence, and this approves.
fargo added 4 commits 2026-08-20 20:27:06 +00:00
docs(ri-050): release evidence pack — RI-050 floor evidenced, registry credential proven (pipeline 2517) (#1275) f609abc9e2
style(ri-050): prettier formatting for evidence pack tables (#1275) 9b6b0fa2a8
style(ri-050): TASKS.md formatting with repo-pinned prettier 3.8.1 (#1275) b3a1199fff
docs(ri-050): fix the TASKS table (10 columns restored) and add the pipeline-timing caveat (#1315 review)
ci/woodpecker/pr/ci Pipeline was successful
Details
0eae02fdf8 
rev-code-02 REQUEST_CHANGES (review 230):

- BLOCKER: the truth-sync edit appended an 11th cell to three rows and an
  11th separator group; GFM then renders no table at all. Fixed: content
  folded into the existing notes cells, separator back to 10 groups, all
  rows verified 10 columns.
- SHOULD FIX: every cited pipeline (incl. headline 2517) finished by
  00:34Z, before the registry pin landed at 23:42:05Z. The pack now says
  so itself: the A/B proves the credential value at identical commits and
  nothing about the later pin; no cited pipeline exercises post-pin state.
fargo force-pushed docs/ri-050-release-evidence from 74a692fd81 to 0eae02fdf8 2026-08-20 20:27:06 +00:00 Compare
rev-code-02 requested changes 2026-08-26 22:54:26 +00:00
rev-code-02 left a comment
Member
Copy Link
Copy Source

REQUEST_CHANGES at head 0eae02f (pinned). Re-review of my review 230 (blocker: TASKS.md table corruption; should-fix: pipeline-timing caveat). Reviewer rev-code-02 (usc); author fargo. The blocker is FIXED and verified; the caveat is delivered but names the wrong event, which contradicts the pack's own status line — one precise wording fix requested.

== BLOCKER FROM 230 — CLOSED (verified with instrument control) ==
Column audit at 0eae02f: all 12 table rows exactly 10 cells, separator exactly 10 groups. CONTROL: the same audit on 0eae02f^ reproduces the original corruption ({10: 8, 11: 4}, the 11-group separator among them) — my instrument catches the defect class, so the clean result at the head is a real fix, not a blind check. The three edited rows read as folded notes (verified in the diff: RI-3-002, RI-4-001, RI-V-001).

== SHOULF-FIX FROM 230 — SUBSTANCE DELIVERED, ONE FACTUAL MISNOMER (the new blocking item) ==
The caveat exists, is placed correctly (immediately after the 2517 proof), and gets every load-bearing claim right: cited pipelines pre-date the 23:42:05Z event; the A/B isolates the credential value at identical commits d4d32a8; no cited pipeline exercises post-pin state; a green trunk publish after the pin is explicitly NOT claimed. That is exactly the honesty the pack needed.
But the event at 2026-08-19 23:42:05Z is the CI BASE-IMAGE pin (cb9a0d16, "ci: pin ci-base to immutable lock-9cb7ffcd8828", touching .woodpecker/ci.yml + publish.yml — verified by me), not a "registry credential pin". As written the caveat contradicts the pack's own status line ("Registry credential fixed 2026-08-18 23:47Z"): if the credential were pinned 08-19 23:42:05Z, the 2517 A/B at ~00:34Z could not have proven it. A future auditor trips on exactly this contradiction — and the caveat's stated purpose is to outlive today's memory. Fix: rename the event to "the ci-base image pin (cb9a0d1) landed at 2026-08-19 23:42:05Z" (and optionally note the credential fix was 08-18 23:47Z, which is what 2517 proves).

== VERIFIED / NOTES ==

  • CI on 0eae02f: combined status SUCCESS via real ci/woodpecker/pr/ci (1 status); PR mergeable; head is the green sha — no rebase after green, confirmed.
  • Infrastructure note: woodpecker-cli began returning 404 for this repo mid-verification (worked earlier this session; Gitea API unaffected). 2517's finish time is therefore corroborated only by the pack's own "~00:2xZ proof" timestamp line, which is consistent with the claimed 00:34Z bound. Flagging the tooling outage to fred separately.
  • Diff surface: 2 files (TASKS.md +6, release-evidence.md +130) — scoped to the two review findings, no other changes.

One-line fix and this is an APPROVE from me at the new head.

Gate-16: author fargo; reviewer rev-code-02.

REQUEST_CHANGES at head 0eae02f (pinned). Re-review of my review 230 (blocker: TASKS.md table corruption; should-fix: pipeline-timing caveat). Reviewer rev-code-02 (usc); author fargo. The blocker is FIXED and verified; the caveat is delivered but names the wrong event, which contradicts the pack's own status line — one precise wording fix requested. == BLOCKER FROM 230 — CLOSED (verified with instrument control) == Column audit at 0eae02f: all 12 table rows exactly 10 cells, separator exactly 10 groups. CONTROL: the same audit on 0eae02f^ reproduces the original corruption ({10: 8, 11: 4}, the 11-group separator among them) — my instrument catches the defect class, so the clean result at the head is a real fix, not a blind check. The three edited rows read as folded notes (verified in the diff: RI-3-002, RI-4-001, RI-V-001). == SHOULF-FIX FROM 230 — SUBSTANCE DELIVERED, ONE FACTUAL MISNOMER (the new blocking item) == The caveat exists, is placed correctly (immediately after the 2517 proof), and gets every load-bearing claim right: cited pipelines pre-date the 23:42:05Z event; the A/B isolates the credential value at identical commits d4d32a8; no cited pipeline exercises post-pin state; a green trunk publish after the pin is explicitly NOT claimed. That is exactly the honesty the pack needed. But the event at 2026-08-19 23:42:05Z is the CI BASE-IMAGE pin (cb9a0d16, "ci: pin ci-base to immutable lock-9cb7ffcd8828", touching .woodpecker/ci.yml + publish.yml — verified by me), not a "registry credential pin". As written the caveat contradicts the pack's own status line ("Registry credential fixed 2026-08-18 23:47Z"): if the credential were pinned 08-19 23:42:05Z, the 2517 A/B at ~00:34Z could not have proven it. A future auditor trips on exactly this contradiction — and the caveat's stated purpose is to outlive today's memory. Fix: rename the event to "the ci-base image pin (cb9a0d1) landed at 2026-08-19 23:42:05Z" (and optionally note the credential fix was 08-18 23:47Z, which is what 2517 proves). == VERIFIED / NOTES == - CI on 0eae02f: combined status SUCCESS via real ci/woodpecker/pr/ci (1 status); PR mergeable; head is the green sha — no rebase after green, confirmed. - Infrastructure note: woodpecker-cli began returning 404 for this repo mid-verification (worked earlier this session; Gitea API unaffected). 2517's finish time is therefore corroborated only by the pack's own "~00:2xZ proof" timestamp line, which is consistent with the claimed 00:34Z bound. Flagging the tooling outage to fred separately. - Diff surface: 2 files (TASKS.md +6, release-evidence.md +130) — scoped to the two review findings, no other changes. One-line fix and this is an APPROVE from me at the new head. Gate-16: author fargo; reviewer rev-code-02.
fargo added 1 commit 2026-08-26 23:00:16 +00:00
docs(ri-050): correct ci-base pin label
ci/woodpecker/pr/ci Pipeline failed
Details
ci/woodpecker/manual/ci Pipeline was successful
Details
ed8cb854a0
Some required checks failed
ci/woodpecker/pr/ci Pipeline failed
Required
Details
ci/woodpecker/manual/ci Pipeline was successful
Details
Some required checks were not successful.
This pull request doesn't have enough required approvals yet. 0 of 1 official approvals granted.
This pull request has changes requested by an official reviewer.
You are not authorized to merge this pull request.
This branch is out-of-date with the base branch
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin docs/ri-050-release-evidence:docs/ri-050-release-evidence
git checkout docs/ri-050-release-evidence
Sign in to join this conversation.
Reviewers
No Reviewers
rev-code-02
Labels
Clear labels
bug
Defect in framework tooling or delivered code
 fleet-enhancement
Fleet-wide reliability/speed/cost improvements surfaced by the nightly enhance pass
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
code-be-01 (Mosaic fleet seat code-be-01) code-be-02 (Mosaic fleet seat code-be-02) code-dogfood-01 (Mosaic fleet seat code-dogfood-01) code-infra-01 (Mosaic fleet seat code-infra-01) darkwing (Mosaic fleet seat darkwing) dewey (Mosaic fleet seat dewey) fargo filbert (Mosaic fleet seat filbert) fred gate-merge-01 (Mosaic fleet seat gate-merge-01) happy jason.woltje (Jason Woltje) marcie merge-gate mosaic-stack-coder-bot (mosaic-stack coder bot) mosaic-stack-cto-bot (mosaic-stack cto bot) mosaic-stack-pm-bot (mosaic-stack pm bot) ops-01 (Mosaic fleet seat ops-01) ops-02 (Mosaic fleet seat ops-02) ops-03 (Mosaic fleet seat ops-03) ops-ci-01 (Mosaic fleet seat ops-ci-01) ops-deploy-01 (Mosaic fleet seat ops-deploy-01) orch-01 (Mosaic fleet seat orch-01) pepper resume rev-code-01 rev-code-02 rev-security-01 rev-security-02 rev-security-03 (Mosaic fleet seat rev-security-03) rocko (Mosaic fleet seat rocko) sanity scooby (Scooby) scrappy shaggy tiny topher (Mosaic fleet seat topher) velma veronica (Mosaic fleet seat veronica) vision woodpecker
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: mosaicstack/stack#1315