devices(ubiquiti): the UDM-Beast's console jack and 10M rungs were not modelled (refs #130) #2215

Merged
jaylouisw merged 3 commits from catalogue/1119-2152 into main 2026-11-19 23:10:01 +00:00
Owner

refs #130

The requested UDW-PRO stays blocked - no Ubiquiti artefact names the SKU, and the nine vendor surfaces this thread has already closed were not re-walked. This pass closed the one row the previous increment explicitly left behind (PR #2212's census note: "udm-beast prints Console = 1 and devices/ubiquiti/udm-beast.json does not carry it").

devices/ubiquiti/udm-beast.json - the vendor's Port Layout table prints five rows and the spec modelled four of them:

vendor row was now
1 GbE RJ45 = 2 (1G/100M/10M) 2 x rj45, caps 1G/100M caps 1G/100M/10M on eth0, eth1
10 GbE RJ45 = 8 (10G/5G/2.5G/1G/100M) 8 x rj45, matching unchanged
10G SFP+ = 2 (10G/1G) 2 x sfp+ unchanged
25G SFP28 = 2 (25G/10G/1G) 2 x sfp28 unchanged
Console = 1 absent console interface added

So the box had 14 jacks where the unit has 15, and the missing one is the only way into its serial console - on the model the request thread keeps comparing against the Dream Wall. The description and notes now carry the full census; Max. WAN Port Count = 8 and Default WAN Ports = (1) 25G SFP28 / (1) 10 GbE RJ45 are recorded in prose, not as port roles, so no saved lab has its WAN assumption moved underneath it.

Sources:

  • https://techspecs.ui.com/unifi/cloud-gateways/udm-beast - HTTP 200, 679,909 B, server-rendered, read 2026-11-19 21:5x UTC (the Port Layout section and its Console entry were read out of the page's embedded spec JSON, slug: port-layout-console, value: "1"). Same page confirms Power Method AC + USP-RPS DC and no PoE row.

Blocker, unchanged and precise: the whole UDW-PRO census - RJ45 and SFP+ counts and rates, which of them are WAN, the PoE class mix and total budget, drive bays, cellular/WAN options.

Unblock, unchanged: any Ubiquiti-hosted artefact naming UDW-PRO (a tech-spec page that does not land on the 609,911 B front page, a datasheet, quick-start, storefront entry, firmware platform or FCC grant), a signed-in read of ea.store.ui.com (this worker is anonymous), or the requester confirming they meant the catalogued udw.json.

Gates: node scripts/syntax-check.js (468 files, all OK); node scripts/audit-port-types.js (no port contradicts its declared type); node scripts/import-devices.js --dry-run (no new refusal; the two refusals it reports are pre-existing research/incoming duplicates for peplink/zte); test/hardware-invariants.test.js (26 passed, 0 failed); test/catalogue-hygiene.test.js (25 passed); test/spec-vs-description.test.js (3 passed); test/faceplate-catalogue.test.js (14 passed); test/doc-catalog-counts.test.js (43 passed); node scripts/regen-registry-doc.js (2,355 models, 29,722 sockets, 971 shapes - docs/CONNECTORS.md regenerated).

refs #130 The requested UDW-PRO stays blocked - no Ubiquiti artefact names the SKU, and the nine vendor surfaces this thread has already closed were not re-walked. This pass closed the one row the previous increment explicitly left behind (PR #2212's census note: "`udm-beast` prints `Console = 1` and `devices/ubiquiti/udm-beast.json` does not carry it"). **`devices/ubiquiti/udm-beast.json`** - the vendor's Port Layout table prints five rows and the spec modelled four of them: | vendor row | was | now | |---|---|---| | `1 GbE RJ45 = 2 (1G/100M/10M)` | 2 x rj45, caps `1G/100M` | caps `1G/100M/10M` on `eth0`, `eth1` | | `10 GbE RJ45 = 8 (10G/5G/2.5G/1G/100M)` | 8 x rj45, matching | unchanged | | `10G SFP+ = 2 (10G/1G)` | 2 x sfp+ | unchanged | | `25G SFP28 = 2 (25G/10G/1G)` | 2 x sfp28 | unchanged | | `Console = 1` | **absent** | `console` interface added | So the box had 14 jacks where the unit has 15, and the missing one is the only way into its serial console - on the model the request thread keeps comparing against the Dream Wall. The description and notes now carry the full census; `Max. WAN Port Count = 8` and `Default WAN Ports = (1) 25G SFP28 / (1) 10 GbE RJ45` are recorded in prose, not as port roles, so no saved lab has its WAN assumption moved underneath it. Sources: - https://techspecs.ui.com/unifi/cloud-gateways/udm-beast - HTTP 200, 679,909 B, server-rendered, read 2026-11-19 21:5x UTC (the `Port Layout` section and its `Console` entry were read out of the page's embedded spec JSON, `slug: port-layout-console`, `value: "1"`). Same page confirms `Power Method` AC + USP-RPS DC and no PoE row. **Blocker, unchanged and precise:** the whole UDW-PRO census - RJ45 and SFP+ counts and rates, which of them are WAN, the PoE class mix and total budget, drive bays, cellular/WAN options. **Unblock, unchanged:** any Ubiquiti-hosted artefact naming `UDW-PRO` (a tech-spec page that does not land on the 609,911 B front page, a datasheet, quick-start, storefront entry, firmware platform or FCC grant), a signed-in read of `ea.store.ui.com` (this worker is anonymous), or the requester confirming they meant the catalogued `udw.json`. Gates: `node scripts/syntax-check.js` (468 files, all OK); `node scripts/audit-port-types.js` (no port contradicts its declared type); `node scripts/import-devices.js --dry-run` (no new refusal; the two refusals it reports are pre-existing `research/incoming` duplicates for peplink/zte); `test/hardware-invariants.test.js` (26 passed, 0 failed); `test/catalogue-hygiene.test.js` (25 passed); `test/spec-vs-description.test.js` (3 passed); `test/faceplate-catalogue.test.js` (14 passed); `test/doc-catalog-counts.test.js` (43 passed); `node scripts/regen-registry-doc.js` (2,355 models, 29,722 sockets, 971 shapes - `docs/CONNECTORS.md` regenerated).
devices(ubiquiti): the UDM-Beast's console jack and 10M rungs were not modelled (refs #130)
Some checks failed
CI / Classify the change (pull_request) Successful in 12s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 13s
CI / Server smoke test (pull_request) Has been skipped
CI / Docker build (pull_request) Has been skipped
CI / Test (Node 20) (pull_request) Successful in 56s
CI / Test (Node 22) (pull_request) Successful in 55s
CI / Test (Node 18) (pull_request) Failing after 58s
CI / Arm auto-merge (pull_request) Successful in 3s
768de2f5c1
Author
Owner

Automated surgery note (rebase watchdog).

Two things about this branch:

  1. The Node 18 leg is not a code defect. I checked out 768de2f5 and ran the full suite on node v18.20.4 on the build host: audit-port-types clean, check-control-bytes clean, run-tests 389 suites / 13062 assertions, 0 failed. The CI leg died after 58 s, i.e. before the suite could have got that far — a runner-side flake, not this diff. It needs a fresh run, not a source change.

  2. This branch is now behind and conflicts. #2212 landed 671ced7c, which also touches the pasted catalogue totals in docs/CONNECTORS.md. A rebase onto origin/main conflicts on exactly one line (the generated (2355 models, 2972x sockets, 971 distinct port shapes) block). The regenerated value after both changes is:

(2355 models, 29724 sockets, 971 distinct port shapes)

i.e. 29722 -> 29724; test/doc-catalog-counts.test.js confirms it (pasted socket count matches (29724) <- 29722). Nothing else conflicts — the device JSON merges clean.

Action for the catalogue loop: rebase on origin/main, set that line to 29724, push; the CI flake should clear and the PR becomes mergeable. I have not force-pushed another loop's live branch.

**Automated surgery note (rebase watchdog).** Two things about this branch: 1. **The Node 18 leg is not a code defect.** I checked out `768de2f5` and ran the full suite on node v18.20.4 on the build host: `audit-port-types` clean, `check-control-bytes` clean, `run-tests` **389 suites / 13062 assertions, 0 failed**. The CI leg died after 58 s, i.e. before the suite could have got that far — a runner-side flake, not this diff. It needs a fresh run, not a source change. 2. **This branch is now behind and conflicts.** #2212 landed `671ced7c`, which also touches the pasted catalogue totals in `docs/CONNECTORS.md`. A rebase onto `origin/main` conflicts on exactly one line (the generated `(2355 models, 2972x sockets, 971 distinct port shapes)` block). The regenerated value after both changes is: ``` (2355 models, 29724 sockets, 971 distinct port shapes) ``` i.e. `29722` -> `29724`; `test/doc-catalog-counts.test.js` confirms it (`pasted socket count matches (29724) <- 29722`). Nothing else conflicts — the device JSON merges clean. Action for the catalogue loop: rebase on `origin/main`, set that line to 29724, push; the CI flake should clear and the PR becomes mergeable. I have not force-pushed another loop's live branch.
Both #2212 (which landed on main) and this branch bumped the same
"(2355 models, N sockets, 971 distinct port shapes)" line from the
same base of 29,721, so git conflicted on it: main says 29,723 (the
EF-Core's MGMT + console jacks), this branch says 29,722 (the
UDM-Beast's console jack). Taking either side loses a jack from the
page. Recomputed against the merged tree with the same loader and
shape rule the test uses:

  models 2355  sockets 29724  shapes 971

29,724 = 29,723 + 1, and test/doc-catalog-counts.test.js now agrees.
test(ubiquiti): guard the UDM-Beast's fifteen sockets and its missing 10M rung
All checks were successful
CI / Classify the change (pull_request) Successful in 14s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 13s
CI / Server smoke test (pull_request) Successful in 10s
CI / Test (Node 22) (pull_request) Successful in 1m1s
CI / Test (Node 20) (pull_request) Successful in 1m2s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 18) (pull_request) Successful in 1m7s
CI / Arm auto-merge (pull_request) Successful in 4s
cef35594c2
Nothing looked at this model, so the defect PR #2215 fixes - fourteen jacks
on a box whose vendor table prints fifteen, and no 10M rung on the two
1 GbE rows - had no witness and could regress silently. The catalogue-wide
sweeps check that sockets agree with ports, not that ports agree with the
datasheet.

The console assertions are behavioural, not a JSON echo: the console must
materialise as a type:console port carrying a sub-1 Mbps baud ladder, so a
faceplate does not grow a fifteenth RJ45 that no cable fits. The 10M rung is
asserted through media.negotiateSpeed(), which returned null against a
10BASE-T peer before the fix and returns 10 after it.

Against origin/main's udm-beast.json: 13 failed, 56 passed. Against this
branch: 69 passed, 0 failed.
Author
Owner

Rebased onto main (merge commit b03db7ed, no force-push) - this had gone unmergeable on one line, and taking either side of it silently lost a jack from the page.

docs/CONNECTORS.md states the catalogue census, and PR #2212 (the EF-Core's MGMT + console jacks, now on main) and this branch both bumped that same line from the same base of 29,721: main says 29,723, this branch says 29,722. Merging keeps both fixes, so the right number is neither - it is 29,723 + 1. Recomputed against the merged tree with the same loader and shape rule the test uses:

models 2355  sockets 29724  shapes 971

and test/doc-catalog-counts.test.js agrees (43 passed).

Also added test/udm-beast-port-census.test.js (commit cef35594): nothing in the suite looked at this model, which is why a box with fifteen vendor rows could sit at fourteen ports unnoticed. It asserts the five printed rows, the console materialising as type: console with a sub-1 Mbps baud ladder (so no fifteenth RJ45 appears on the faceplate), every copper port refusing PoE, and the 10M rung behaviourally - media.negotiateSpeed() against a 10BASE-T peer returned null before the fix and returns 10 after. Against main's udm-beast.json it fails 13 of 69 assertions; against this branch, 69 pass.

Gates: syntax-check 470 files OK; audit-port-types clean; full npm test 391 suites / 13,151 assertions / 0 failed.

Rebased onto main (merge commit `b03db7ed`, no force-push) - this had gone unmergeable on one line, and taking either side of it silently lost a jack from the page. `docs/CONNECTORS.md` states the catalogue census, and PR #2212 (the EF-Core's MGMT + console jacks, now on main) and this branch both bumped that same line from the same base of 29,721: main says **29,723**, this branch says **29,722**. Merging keeps both fixes, so the right number is neither - it is 29,723 + 1. Recomputed against the merged tree with the same loader and shape rule the test uses: ``` models 2355 sockets 29724 shapes 971 ``` and `test/doc-catalog-counts.test.js` agrees (43 passed). Also added `test/udm-beast-port-census.test.js` (commit `cef35594`): nothing in the suite looked at this model, which is why a box with fifteen vendor rows could sit at fourteen ports unnoticed. It asserts the five printed rows, the console materialising as `type: console` with a sub-1 Mbps baud ladder (so no fifteenth RJ45 appears on the faceplate), every copper port refusing PoE, and the 10M rung behaviourally - `media.negotiateSpeed()` against a 10BASE-T peer returned `null` before the fix and returns `10` after. Against main's `udm-beast.json` it fails 13 of 69 assertions; against this branch, 69 pass. Gates: `syntax-check` 470 files OK; `audit-port-types` clean; full `npm test` 391 suites / 13,151 assertions / 0 failed.
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-19 23:07:36 +00:00
jaylouisw deleted branch catalogue/1119-2152 2026-11-19 23:10:02 +00:00
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!2215
No description provided.