[gunicsba]

Summary

Section control detection now looks only at process data that a device element actually links. The method and its targets are resolved once, when the implement activates, by a pure function, resolve_section_control(). A new test runs that function against every sample DDOP.

Draft for review. Some of these paths have only synthetic evidence (see Evidence below), so part of the discussion is whether to keep them.

Commits

1. Add sample implement DDOPs for tests: the 24 real pools plus manifest.csv and README from the integration build, unchanged. They aren't on develop yet.
2. Add synthetic MIDI 3 section control variants: three *_synthetic.ddop files, derived from the real MIDI 3 capture by the firmware's own DDOP rules (from reverse engineering). The same rules rebuild the real capture byte for byte. The manifest and README mark them as never seen on the wire.
3. Resolve section control targets from linked process data only: the refactor, the fix and the test.

What changes in the TC

• Process data that no element references is ignored. The MIDI 3 sends every template object, so its pool declares DDI 289, 291–305 and 162–176 objects that belong to no element.
• "Settable" is checked on the object a given element links, not on the first object in the pool with that DDI. is_ddi_settable() is removed.
• New per-element method: DDI 289 on each section, actual state on DDI 141 of the same section. Settable DDI 141 stays as the fallback.
• The global DDI 289 is only taken from an element that is not a section.
• "DDI 289 not available" is logged once per client at activation, not on every send.
• Device element children are read by index, because iterating get_child_object_ids() reads past the end of the list (Bump AgIsoStack to pick up the DataSpan fix (wrong DDI-to-element mapping) #110).

Order of preference: DDI 290 > settable DDI 161 > per-element DDI 289 > per-element settable DDI 141. This follows the fallback rules in the ISO 11783-11 attachment for DDI 289/290: a TC "shall be able to fall back to using the Actual Work State DDIs" when the implement marks them settable.

Evidence

Change	Evidence
Child IDs read by index	Real. On the real MIDI 3 capture, the old logic mapped DDI 291 to element 15 and DDI 289 to random sections, and DDI 290 on PWRW_II to element 17. The result changed between runs. Same root cause as #109.
DDI 289 log once	Real. The MIDI 3 and SKY Easydrill pools have no linked DDI 289.
Only linked process data counts	Synthetic. Every real pool already resolves correctly without it. It matters for the MIDI 3 in old-SC or per-element mode.
Settable check per linked object	Synthetic. It matters for MIDI 3 old-SC + condensed mode, where the unlinked non-settable DDI 161 comes first in the pool.
Per-element DDI 289	Synthetic and the most speculative. The data dictionary allows the layout, but the same attachment makes condensed 290/161 mandatory for AEF TC-SC.

If the synthetic-only paths aren't wanted, the per-element DDI 289 method is the easiest to drop: one branch in the resolver and one manifest row.

Test results

ctest: 6 of 7 pass.

• Before the fix, using today's logic moved into the resolver unchanged, the three synthetic MIDI 3 files fail as predicted.
• After the fix, all four MIDI 3 pools resolve as follows:
  • capture: 290 on element 2
  • old SC, condensed: 161 on element 2
  • new SC, per-element: 289 on elements 10–20
  • old SC, per-element: 141 on elements 10–20
• Every other pool keeps the method the TC picks today.

section_control_method still fails on 4 manifest rows, to be decided before merge. The Kuhn Espro DDI141 files (en and hu) and the JD 6000M and 6R175 tractors are listed as DDI141, but their DDI 141 is not settable. Today's TC and this branch both pick none for them (issue #10: "SC switch is not working"). The proposal is to change those rows to none; no code change is needed.

Notes for reviewers

• Conflicts with Keep implements in automatic mode on AOG's current section demand #106 (draft), which edits the same lines in send_section_setpoint_states().
• The child-list loops in request_measurement_commands() still iterate the broken span; Bump AgIsoStack to pick up the DataSpan fix (wrong DDI-to-element mapping) #110 fixes those.
• Behaviour change: the global DDI 289 is now also sent in settable-161 mode if the implement has one. By rule 4 of the attachment that combination shouldn't occur.
• The executable target is test_section_control, so cmake-format and cmake-lint agree on the wrapping.

Not tested

• Real implements: no MIDI 3 is configured in old-SC or per-element mode.
• Only Windows (MSVC Release) was built locally.

🤖 Generated with Claude Code