devices(ubiquiti): the Express 7 is a router, and it was filed as an access point (refs #130) #2156

Merged
jaylouisw merged 1 commit from catalogue/1119-1252 into main 2026-11-19 14:27:18 +00:00
Owner

refs #130 — this does not close the Dream Wall Pro request (still blocked, unchanged below). It settles the other half of the stated lab: a router/gateway node for a routed lab with NAT and DHCP.

What was wrong

devices/ubiquiti/ux7.json described the UniFi Express 7 as category: "wireless" with wireless, l2-forward, cli, http-server, dhcp-client. The engine branches on capabilities, so the Express 7 could not route, could not NAT, could not hand out a lease and could not filter — and because the palette filters on category, choosing "Router" did not offer it at all. Nothing failed, because an access point that sits on a network looks exactly like a gateway that does nothing.

field was now vendor row
category wireless router filed under Cloud Gateways; BSSID row reads "4 per Radio (Gateway Mode) / 4 per Radio (AP Mode)"
l3-forward absent present "Dynamic Routing — OSPF ✓"
nat absent present "Advanced NAT (SNAT / DNAT / Masquerade / NAT Pooling / 1-to-1 NAT) ✓"
dhcp-server absent present "Customizable DHCP Server ✓"
firewall absent present "Stateful Firewall ✓", "VLAN/Subnet-based Traffic Segmentation ✓"
eth0 speedCaps 2.5G/1G/100M + 10M "2.5 GbE RJ45 1 (2.5G/1G/100M/10M)"
radios 2 × 1G 688M / 4.3G / 5.7G Throughput: 2.4 GHz 688 Mbps, 5 GHz 4.3 Gbps, 6 GHz 5.7 Gbps, 2x2 each
description "WiFi 6/7" "WiFi 6/6E + 7", default WAN named "Default WAN Ports (1) 10 GbE RJ45"

Port count is unchanged (1x 2.5G RJ45 + 1x 10G RJ45); the 10G port is the vendor's default WAN, recorded in the description and the notes rather than as a role, so no saved lab changes which port it treats as uplink.

Sources (read 2026-11-19, crawler UA)

  • https://techspecs.ui.com/unifi/cloud-gateways/ux7 — HTTP 200, 688,360 B, title "UniFi Express 7 - Tech Specs" (not the 609,911 B techspecs home page that means "no such slug"). Rows quoted above; also "Max. Power Consumption 22W", "Power Method USB type C (5V DC/5A)".
  • https://store.ui.com/us/en/products/ux7 — HTTP 200, 587,404 B, SKU UX7-2: "Mesh-scalable, super-compact 10G Cloud Gateway with integrated 6 GHz WiFi 7".
  • https://dl.ui.com/qig/ux7 — the tech-spec page lists it as the only document; it serves an HTML viewer (7,062 B), not a text-layer PDF, so nothing was read from it.

No datasheet exists for this SKU, so the rendered spec table is the vendor source — same position as udr7.json.

Test fix that the data change exposed

test/port-roundtrip.test.js looked for a port whose PoE the engine inferred inside specs.slice(0, 400). The only such port in the catalogue sits at index 399 (aruba/2530-8-PoE+ port 1); moving UX7 into the router bucket pushed it to 400 and the suite failed with "no such port found" while no PoE field had changed. The loop already breaks on its first candidate, so scanning the whole catalogue costs the same 4.6 s. The assertion ("exactly one candidate, and it does not start claiming it declared the wattage") is unchanged — the window was the bug.

Gates

  • node scripts/syntax-check.js — 453 files, all OK
  • node scripts/audit-port-types.js — no port contradicts its declared type
  • node scripts/audit-poe-budget.js — 0 budgets with no PSE port
  • node scripts/check-control-bytes.js — none
  • node scripts/check-doc-citations.js — 358 citations, none judged stale
  • node scripts/regen-registry-doc.js — regenerated, idempotent (2,356 models / 29,717 sockets / 972 shapes)
  • node scripts/run-tests.js — 374 suites, 12,708 assertions, 0 failed (includes hardware-invariants 26/0, spec-vs-description, importer tripwire)

Still blocked on #130

The Dream Wall Pro has no Ubiquiti surface at all (techspecs, dl.ui.com, qig/, storefronts, grantee SWX, firmware platform index, controller device DB, Wayback CDX), so no UDW-PRO spec was written and none of the UDW's numbers were used. Unchanged: RJ45/SFP+ counts and rates, which are WAN, PoE class mix and budget, drive bays, cellular/WAN options — or a note that the requester meant the UDW.

refs #130 — this does **not** close the Dream Wall Pro request (still blocked, unchanged below). It settles the other half of the stated lab: a router/gateway node for a routed lab with NAT and DHCP. ## What was wrong `devices/ubiquiti/ux7.json` described the UniFi Express 7 as `category: "wireless"` with `wireless, l2-forward, cli, http-server, dhcp-client`. The engine branches on `capabilities`, so the Express 7 could not route, could not NAT, could not hand out a lease and could not filter — and because the palette filters on category, choosing "Router" did not offer it at all. Nothing failed, because an access point that sits on a network looks exactly like a gateway that does nothing. | field | was | now | vendor row | |---|---|---|---| | category | `wireless` | `router` | filed under Cloud Gateways; BSSID row reads "4 per Radio (Gateway Mode) / 4 per Radio (AP Mode)" | | l3-forward | absent | present | "Dynamic Routing — OSPF ✓" | | nat | absent | present | "Advanced NAT (SNAT / DNAT / Masquerade / NAT Pooling / 1-to-1 NAT) ✓" | | dhcp-server | absent | present | "Customizable DHCP Server ✓" | | firewall | absent | present | "Stateful Firewall ✓", "VLAN/Subnet-based Traffic Segmentation ✓" | | eth0 speedCaps | 2.5G/1G/100M | + 10M | "2.5 GbE RJ45 1 (2.5G/1G/100M/10M)" | | radios | 2 × `1G` | 688M / 4.3G / 5.7G | Throughput: 2.4 GHz 688 Mbps, 5 GHz 4.3 Gbps, 6 GHz 5.7 Gbps, 2x2 each | | description | "WiFi 6/7" | "WiFi 6/6E + 7", default WAN named | "Default WAN Ports (1) 10 GbE RJ45" | Port count is unchanged (1x 2.5G RJ45 + 1x 10G RJ45); the 10G port is the vendor's default WAN, recorded in the description and the `notes` rather than as a `role`, so no saved lab changes which port it treats as uplink. ## Sources (read 2026-11-19, crawler UA) - https://techspecs.ui.com/unifi/cloud-gateways/ux7 — HTTP 200, 688,360 B, title "UniFi Express 7 - Tech Specs" (not the 609,911 B techspecs home page that means "no such slug"). Rows quoted above; also "Max. Power Consumption 22W", "Power Method USB type C (5V DC/5A)". - https://store.ui.com/us/en/products/ux7 — HTTP 200, 587,404 B, SKU `UX7-2`: "Mesh-scalable, super-compact 10G Cloud Gateway with integrated 6 GHz WiFi 7". - https://dl.ui.com/qig/ux7 — the tech-spec page lists it as the only document; it serves an HTML viewer (7,062 B), not a text-layer PDF, so nothing was read from it. No datasheet exists for this SKU, so the rendered spec table is the vendor source — same position as `udr7.json`. ## Test fix that the data change exposed `test/port-roundtrip.test.js` looked for a port whose PoE the engine inferred inside `specs.slice(0, 400)`. The only such port in the catalogue sits at index 399 (`aruba/2530-8-PoE+` port 1); moving UX7 into the `router` bucket pushed it to 400 and the suite failed with "no such port found" while no PoE field had changed. The loop already breaks on its first candidate, so scanning the whole catalogue costs the same 4.6 s. The assertion ("exactly one candidate, and it does not start claiming it declared the wattage") is unchanged — the window was the bug. ## Gates - `node scripts/syntax-check.js` — 453 files, all OK - `node scripts/audit-port-types.js` — no port contradicts its declared type - `node scripts/audit-poe-budget.js` — 0 budgets with no PSE port - `node scripts/check-control-bytes.js` — none - `node scripts/check-doc-citations.js` — 358 citations, none judged stale - `node scripts/regen-registry-doc.js` — regenerated, idempotent (2,356 models / 29,717 sockets / 972 shapes) - `node scripts/run-tests.js` — 374 suites, 12,708 assertions, 0 failed (includes hardware-invariants 26/0, spec-vs-description, importer tripwire) ## Still blocked on #130 The Dream Wall Pro has no Ubiquiti surface at all (techspecs, `dl.ui.com`, `qig/`, storefronts, grantee SWX, firmware platform index, controller device DB, Wayback CDX), so no UDW-PRO spec was written and none of the UDW's numbers were used. Unchanged: RJ45/SFP+ counts and rates, which are WAN, PoE class mix and budget, drive bays, cellular/WAN options — or a note that the requester meant the UDW.
devices(ubiquiti): the Express 7 is a router, and it was filed as an access point (refs #130)
Some checks failed
CI / Classify the change (pull_request) Successful in 10s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Server smoke test (pull_request) Successful in 9s
CI / Test (Node 20) (pull_request) Failing after 56s
CI / Test (Node 22) (pull_request) Successful in 57s
CI / Docker build (pull_request) Successful in 23s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Arm auto-merge (pull_request) Successful in 3s
4eb6223f15
UX7 declared neither l3-forward, nat, dhcp-server nor firewall, so a lab
asking it for the routed node in #130 got an access point: no NAT, no DHCP
server, no firewall, and the palette filter for "Router" hid it. Ubiquiti
files the SKU under Cloud Gateways and its own tech-spec page ticks
Stateful Firewall, Advanced NAT, Customizable DHCP Server and OSPF.

Also: two radios at "1G" became the three the vendor publishes (6 GHz
5.7, 5 GHz 4.3, 2.4 GHz 688 Mbps), and eth0 gained the 10M rung.

The data change moved UX7 into an earlier category bucket, which pushed
the only port the engine infers PoE for from index 399 to 400 in
test/port-roundtrip.test.js and failed it with "no such port found". The
window was the bug, not the assertion: the loop breaks on its first
candidate, so scanning the whole catalogue costs the same 4.6 s.
Author
Owner

Surgery check (2026-11-19 13:1xZ) — the Node 20 red does not reproduce on this head (4eb6223f).

What I verified: merged origin/main (053b3995) into this branch locally — clean, no conflicts, no add/add. Then ran the full suite under Node v20.18.1 (the dist used by the failing leg): 374 suite(s) … 12708 assertions passed, 0 failed. Node 18 and Node 22 legs on this SHA are green; Node 26 locally is green too.

So the single failing Test (Node 20) leg looks like a flake or a runner-side failure, not a defect in ux7.json/port-roundtrip.test.js. (The slice(0,400) → full-list change in test/port-roundtrip.test.js is a good order-independence fix and is not the cause: it only widens the candidate search.)

Not merging: head-SHA status is failure, and the red-PR rule forbids it. Firmware loop: re-run the Node 20 job, or push any trivial commit to re-trigger the workflow — on a green run this merges clean (I have it stacked on current main, no conflicts).

**Surgery check (2026-11-19 13:1xZ) — the Node 20 red does not reproduce on this head (`4eb6223f`).** What I verified: merged `origin/main` (`053b3995`) into this branch locally — clean, no conflicts, no add/add. Then ran the full suite under **Node v20.18.1** (the dist used by the failing leg): `374 suite(s) … 12708 assertions passed, 0 failed`. Node 18 and Node 22 legs on this SHA are green; Node 26 locally is green too. So the single failing `Test (Node 20)` leg looks like a flake or a runner-side failure, not a defect in `ux7.json`/`port-roundtrip.test.js`. (The `slice(0,400)` → full-list change in `test/port-roundtrip.test.js` is a good order-independence fix and is not the cause: it only widens the candidate search.) Not merging: head-SHA status is `failure`, and the red-PR rule forbids it. Firmware loop: re-run the Node 20 job, or push any trivial commit to re-trigger the workflow — on a green run this merges clean (I have it stacked on current main, no conflicts).
Author
Owner

Rebased onto main (87bbc699) as cf8f8012 — the push is the rebase, no source edits beyond one generated line.

Conflict was mechanical: docs/CONNECTORS.md pasted faceplate-test output, (... 29718 sockets ...) on main vs 29717 on the branch. Main's 29718 already included #2159's added 5G port, so neither side was right for the merged tree. Resolved by measuring: test/doc-catalog-counts.test.js reports the merged catalogue at 2356 models / 29719 sockets / 972 shapes (the ux7 role change nets +1 socket over main), and the line now says 29719. docs/DEVICE_REGISTRY.md came from the generator, untouched by hand.

Full suite on the rebased tree, Node 26: 375 suites / 12750 assertions, 0 failed. The previous red was Test (Node 20) only, with the other two legs green and no failure text retrievable (log endpoint 404s) — that pattern is not reproducible here, so this push exists to get a clean run on a fresh head.

Rebased onto main (87bbc699) as `cf8f8012` — the push is the rebase, no source edits beyond one generated line. Conflict was mechanical: `docs/CONNECTORS.md` pasted faceplate-test output, `(... 29718 sockets ...)` on main vs `29717` on the branch. Main's 29718 already included #2159's added 5G port, so neither side was right for the merged tree. Resolved by measuring: `test/doc-catalog-counts.test.js` reports the merged catalogue at **2356 models / 29719 sockets / 972 shapes** (the ux7 role change nets +1 socket over main), and the line now says 29719. `docs/DEVICE_REGISTRY.md` came from the generator, untouched by hand. Full suite on the rebased tree, Node 26: **375 suites / 12750 assertions, 0 failed**. The previous red was `Test (Node 20)` only, with the other two legs green and no failure text retrievable (log endpoint 404s) — that pattern is not reproducible here, so this push exists to get a clean run on a fresh head.
jaylouisw force-pushed catalogue/1119-1252 from 4eb6223f15
Some checks failed
CI / Classify the change (pull_request) Successful in 10s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Server smoke test (pull_request) Successful in 9s
CI / Test (Node 20) (pull_request) Failing after 56s
CI / Test (Node 22) (pull_request) Successful in 57s
CI / Docker build (pull_request) Successful in 23s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Arm auto-merge (pull_request) Successful in 3s
to cf8f801268
All checks were successful
CI / Classify the change (pull_request) Successful in 12s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 10s
CI / Server smoke test (pull_request) Successful in 9s
CI / Test (Node 20) (pull_request) Successful in 56s
CI / Test (Node 22) (pull_request) Successful in 54s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 18) (pull_request) Successful in 1m1s
CI / Arm auto-merge (pull_request) Successful in 3s
2026-11-19 14:22:08 +00:00
Compare
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-19 14:23:18 +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!2156
No description provided.