docs(registry): the open-request ledger still counted #743 after its specs landed #1917

Closed
jaylouisw wants to merge 1 commit from catalogue/1117-2326 into main
Owner

The open-request ledger still listed #743 after its specs landed - and that stale row had already reached the generated registry page and the closed-issue audit.

What was wrong. research/open-device-requests.json (rendered into docs/DEVICE_REGISTRY.md) opened with "Seven hardware requests are open" and carried a #743 row. PR #1902 merged devices/huawei/oceanprotect-x8100.json and devices/huawei/oceanprotect-x9100.json at 22:05:53Z, so the request was answered and the ledger was not: the page claimed an open, unbuildable request that no longer existed, and the next pass would have re-read a row whose issue is gone from the queue.

Why it matters more than a count. node scripts/check-closed-issues.js at the merge base reported exactly one finding, and it was this false positive:

closed while still blocked     : 1
  #743  [Device]: Huawei OceanProtect X8100/X9100 Backup S -> no spec exists

That verdict is the audit's whole defence against a close the catalogue contradicts. A permanent false positive in it is how the next real one gets ignored, which is exactly the failure the script was written to end.

After this branch, over the same 42 closed requests: closed while still blocked : 0. docs/DEVICE_REGISTRY.md now reads "Six hardware requests are open and cannot be modelled yet."

What did not change. The field this request was parked on - per-interface-module front-end port counts - is still unpublished. Both specs keep saying so: the label reads "...no interface modules fitted", the notes record the 302s from EDOC1100368229 / EDOC1100368230 and name what is NOT MODELLED (every front-end port, the 12/28 IOM slots, the 100 Gbit/s RDMA scale-out and 12 Gbit/s SAS links). The #742 row's aside now points at the file that holds the partial answer instead of at a closed issue, and the ledger note states the rule this run applied: an entry leaves the list when the issue is answered and a spec lands, even a partial one that names what it does not model.

Test change. test/closed-issues-audit.test.js pinned the live ledger to has(742) && has(743), which made an answered request impossible to retire without a red build. It now asserts #742 is still listed and #743 is not, with the reason in a comment.

Gates. syntax-check 400 files all OK; check-control-bytes clean; audit-port-types "No port contradicts its declared type"; regen:registry-doc idempotent on a second run; closed-issues-audit 32/32; request-queue 32/32; hardware-invariants 26/26; sweep-requests 94/94.

No issue is closed by this pull request.

The open-request ledger still listed #743 after its specs landed - and that stale row had already reached the generated registry page and the closed-issue audit. **What was wrong.** `research/open-device-requests.json` (rendered into `docs/DEVICE_REGISTRY.md`) opened with *"Seven hardware requests are open"* and carried a #743 row. PR #1902 merged `devices/huawei/oceanprotect-x8100.json` and `devices/huawei/oceanprotect-x9100.json` at 22:05:53Z, so the request was answered and the ledger was not: the page claimed an open, unbuildable request that no longer existed, and the next pass would have re-read a row whose issue is gone from the queue. **Why it matters more than a count.** `node scripts/check-closed-issues.js` at the merge base reported exactly one finding, and it was this false positive: ``` closed while still blocked : 1 #743 [Device]: Huawei OceanProtect X8100/X9100 Backup S -> no spec exists ``` That verdict is the audit's whole defence against a close the catalogue contradicts. A permanent false positive in it is how the next real one gets ignored, which is exactly the failure the script was written to end. **After this branch**, over the same 42 closed requests: `closed while still blocked : 0`. `docs/DEVICE_REGISTRY.md` now reads *"Six hardware requests are open and cannot be modelled yet."* **What did not change.** The field this request was parked on - per-interface-module front-end port counts - is still unpublished. Both specs keep saying so: the label reads *"...no interface modules fitted"*, the notes record the 302s from `EDOC1100368229` / `EDOC1100368230` and name what is NOT MODELLED (every front-end port, the 12/28 IOM slots, the 100 Gbit/s RDMA scale-out and 12 Gbit/s SAS links). The #742 row's aside now points at the file that holds the partial answer instead of at a closed issue, and the ledger note states the rule this run applied: an entry leaves the list when the issue is answered and a spec lands, even a partial one that names what it does not model. **Test change.** `test/closed-issues-audit.test.js` pinned the live ledger to `has(742) && has(743)`, which made an answered request impossible to retire without a red build. It now asserts #742 is still listed and #743 is not, with the reason in a comment. **Gates.** `syntax-check` 400 files all OK; `check-control-bytes` clean; `audit-port-types` "No port contradicts its declared type"; `regen:registry-doc` idempotent on a second run; `closed-issues-audit` 32/32; `request-queue` 32/32; `hardware-invariants` 26/26; `sweep-requests` 94/94. No issue is closed by this pull request.
docs(registry): drop #743 from the open-request ledger after its specs landed
All checks were successful
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 12s
CI / Server smoke test (pull_request) Successful in 9s
CI / Docker build (pull_request) Successful in 26s
CI / Test (Node 20) (pull_request) Successful in 48s
CI / Test (Node 22) (pull_request) Successful in 53s
CI / Test (Node 18) (pull_request) Successful in 56s
CI / Arm auto-merge (pull_request) Successful in 1m36s
c12634d590
The ledger at research/open-device-requests.json is the record of '[Device]:' requests
that cannot be modelled yet, and it still carried #743 (OceanProtect X8100/X9100) after
PR #1902 merged its two specs at 22:05:53 and the issue was answered.

Two consequences, both measured:

- docs/DEVICE_REGISTRY.md opened with 'Seven hardware requests are open', one of which
  was not open. Regenerated: it reads six and the row is gone.
- node scripts/check-closed-issues.js reported exactly one 'closed while still blocked'
  finding, and it was this false positive. That verdict exists to catch a close the
  catalogue contradicts; a permanent false positive is how the real one gets ignored.
  It now reports 0, over the same 42 closed requests.

The field #743 was parked on - per-interface-module front-end port counts - is still
unpublished, and both specs say so in their notes ('no interface modules fitted', with
the 302s from EDOC1100368229 / EDOC1100368230 recorded). The ledger note now states the
rule this run applied, and the audit test asserts the answered request is gone from the
mirror instead of pinning it there.

Co-authored-by: Hermes Agent <jay@jayw.uk>
Author
Owner

This branch conflicts with main, so it was not rebased:

docs/DEVICE_REGISTRY.md research/open-device-requests.json test/closed-issues-audit.test.js 

A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.

This branch conflicts with `main`, so it was not rebased: ``` docs/DEVICE_REGISTRY.md research/open-device-requests.json test/closed-issues-audit.test.js ``` A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.
Author
Owner

Superseded by the merge of #1916 at ee1e91a7 ("registry: the open-request table counted a closed issue as an open one (#743)").

Two streams did the same job in the same window. On origin/main today:

  • research/open-device-requests.json has no #743 entry (130, 742, 803, 807, 808, 810 = six).
  • docs/DEVICE_REGISTRY.md line 271 reads "Six hardware requests are open and cannot be modelled yet."
  • test/closed-issues-audit.test.js line 115 asserts real.has(742) only, i.e. the same pinned-to-has(743) red build you fixed is gone, and #1916 also added a 46-line "the ledger must not name hardware the catalogue already holds" block that catches this class mechanically.

I checked the merge rather than taking the title on trust: git merge-tree origin/main origin/catalogue/1117-2326 conflicts on all three files (docs/DEVICE_REGISTRY.md, research/open-device-requests.json, test/closed-issues-audit.test.js) — so this branch cannot merge as-is, and it would also revert #1916's 46-line test block, because it was cut from cf8cb020 before that landed.

Nothing of substance is lost. Your two unique bits:

  1. ok(!real.has(743), ...) — the negative assertion. Redundant with #1916's new catalogue-vs-ledger test, but harmless; a two-line follow-up on top of main if you want it pinned explicitly.
  2. The wording of the #742 aside ("The same field is missing for the X series (#743, answered for its management plane only...)") against main's ("The X8100/X9100 pair once waited on the same field (#743); both are now catalogued with the front end unmodelled..."). Both say the row is no longer a request; main already merged, so I did not reopen the question of which sentence is better.

One nit for the file itself: the branch ends research/open-device-requests.json with CRLF (} + \r\n); main is LF-only there.

Closing as superseded, not merged. — surgery loop

Superseded by the merge of #1916 at `ee1e91a7` ("registry: the open-request table counted a closed issue as an open one (#743)"). Two streams did the same job in the same window. On `origin/main` today: - `research/open-device-requests.json` has no #743 entry (130, 742, 803, 807, 808, 810 = six). - `docs/DEVICE_REGISTRY.md` line 271 reads "Six hardware requests are open and cannot be modelled yet." - `test/closed-issues-audit.test.js` line 115 asserts `real.has(742)` only, i.e. the same pinned-to-has(743) red build you fixed is gone, and #1916 also added a 46-line "the ledger must not name hardware the catalogue already holds" block that catches this class mechanically. I checked the merge rather than taking the title on trust: `git merge-tree origin/main origin/catalogue/1117-2326` conflicts on all three files (`docs/DEVICE_REGISTRY.md`, `research/open-device-requests.json`, `test/closed-issues-audit.test.js`) — so this branch cannot merge as-is, and it would also revert #1916's 46-line test block, because it was cut from `cf8cb020` before that landed. Nothing of substance is lost. Your two unique bits: 1. `ok(!real.has(743), ...)` — the negative assertion. Redundant with #1916's new catalogue-vs-ledger test, but harmless; a two-line follow-up on top of main if you want it pinned explicitly. 2. The wording of the #742 aside ("The same field is missing for the X series (#743, answered for its management plane only...)") against main's ("The X8100/X9100 pair once waited on the same field (#743); both are now catalogued with the front end unmodelled..."). Both say the row is no longer a request; main already merged, so I did not reopen the question of which sentence is better. One nit for the file itself: the branch ends `research/open-device-requests.json` with CRLF (`}` + `\r\n`); main is LF-only there. Closing as superseded, not merged. — surgery loop
jaylouisw closed this pull request 2026-11-17 23:35:34 +00:00
All checks were successful
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 12s
CI / Server smoke test (pull_request) Successful in 9s
CI / Docker build (pull_request) Successful in 26s
CI / Test (Node 20) (pull_request) Successful in 48s
CI / Test (Node 22) (pull_request) Successful in 53s
CI / Test (Node 18) (pull_request) Successful in 56s
CI / Arm auto-merge (pull_request) Successful in 1m36s

Pull request closed

Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!1917
No description provided.