[nflyernz]

F17: ArduPlane GNSS serial-port defaults

The AMC 5.0.0 ArduPlane Basic Mandatory Configuration walkthrough found that both generic Plane GNSS templates unconditionally set SERIAL3_PROTOCOL=5. That can change a port without a physical GPS receiver. Serial protocol 5 selects the GPS transport; it is distinct from GPS1_TYPE=5, the NMEA driver.

Remove that hardcoded line from empty_4.7.x/18_gnss.param and normal_plane/18_gnss.param. Plane Step 18 now derives SERIAL1_PROTOCOL through SERIAL9_PROTOCOL as 5 only for the selected UART and a physical serial GPS driver. The firmware-dependent GPS_TYPE and GPS1_TYPE rules are unchanged. The new change reason was added to the generated configuration-step translation strings.

The rules make no serial assignment for None, DroneCAN, MAVLink, MSP, HIL, or ExternalAHRS. No baud rate is inferred.

Validation

• Focused pytest: 83 passed (Plane GNSS regression, configuration-step schema, configuration-strings BDD).
• Ruff check and format, focused Pylint (10.00/10), targeted ty check, and git diff --check: passed.
• Regression tests cover both templates; physical GPS on SERIAL2/3/4; nonserial and special transports; both firmware driver rules; saved serial values and eligible FC import; connection rename; and display of a newly derived parameter after step load.

Inline editor follow-up and dependency

Commit 928697ad11e3e6f fixes the GNSS inline rename exception. Forced, derived, and F17 rule-owned serial destinations are left to their owning rules. Regression tests cover saved GNSS port changes, special transports, repeated changes, and final export.

Validation: 432 focused tests passed, with Ruff, focused Pylint, targeted ty, and git diff --check passing.

Remaining dependency: Changing from None to a physical GNSS UART while Step 18 is open requires the shared inline derived-parameter refresh correction in #2289. Until integrated, reloading the step makes the newly derived parameter available.

No physical flight controller was used. Leave this PR in draft for upstream CI and automated review.

[AP-Review]

Deprecated — see below for the updated review.

Previous review (2026-10-27)

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: REQUEST CHANGES

Reviewed at head 3b807d0b8a.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/ardupilot/methodicconfigurator/2288/1.html#prMethodicConfigurator-2288

Requesting changes: the new derived GNSS serial rules are correct when evaluated, but they break the inline GNSS connection edit on step 18.

Blocking

1. configuration_steps_ArduPlane.json:598: projects whose 18_gnss.param already has a SERIALn_PROTOCOL now raise ParameterForcedOrDerivedError when the GNSS port is changed in the inline editor. That covers every project made from the old Plane templates, and new projects once step 18 has been saved. The cause is that _apply_gnss_protocol_value calls set_new_value on the now-derived rename target. The table is not refreshed and the rename is left half-done. Skip forced/derived targets there, since the rule already gives 5, and add an inline-refresh test with a saved SERIAL3_PROTOCOL.

Should fix

• empty_4.7.x/18_gnss.param:7 (same for normal_plane): with the template row gone, an inline port change from SERIAL3 to SERIAL4 keeps and saves SERIAL3_PROTOCOL=5 and never adds SERIAL4_PROTOCOL. Base renamed it correctly. refresh_current_step_computed_parameters needs to add newly derived rows and drop stale added ones, or the template should keep a row for the rename to move.

[AP-Review]

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: COMMENT

Reviewed at head 928697ad11.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/ardupilot/methodicconfigurator/2288/2.html#prMethodicConfigurator-2288

No blockers remain; one should-fix depends on #2289 landing first.

Since the last review, the ParameterForcedOrDerivedError on inline GNSS port changes with a saved SERIALn_PROTOCOL is fixed by the new guard, and tests now cover it.

Should fix

• empty_4.7.x/18_gnss.param:7 (also normal_plane), template row removal (still open): in a new project, an inline SERIAL3 to SERIAL4 change still saves and uploads SERIAL3_PROTOCOL=5 and no SERIAL4_PROTOCOL, and None to SERIAL4 gives no serial row until the step is reloaded. Merging fix: reconcile inline derived parameters after component edits #2289 onto this head fixes both, so land fix: reconcile inline derived parameters after component edits #2289 first or together, and add a new-template inline port-change test in whichever merges last.

Minor (1)

• configuration_steps_ArduPlane.json:600: with a saved SERIAL3_PROTOCOL, switching the driver inline from uBlox to MAVLink on the same port leaves the row derived, read-only and at 5 until reload. The inline refresh should release a derived row once its rule stops applying. fix: reconcile inline derived parameters after component edits #2289 does not cover this.