[gunicsba]

Summary

DeviceElementObject::get_child_object_ids() returns a DataSpan<const std::uint16_t>. AgIsoStack's DataSpan multiplies by sizeof(T) in end() and operator[], on top of the pointer arithmetic that already scales by sizeof(T):

T *end() const { return ptr + _size * sizeof(T); }   // utility/include/isobus/utility/data_span.hpp

For std::uint16_t, iterating it reads twice as many IDs as the element has children. The second half comes from past the end of the list, so whatever is in that memory can match an object ID. Checked with AgIsoStack's own deserializer: an element with 22 children yields 44 values, the last 22 of them junk.

All four TC loops that map DDIs to elements iterate this span (find_owning_element_number, the section parent lookup, and both measurement-subscription loops).

Field symptom

Local bench test with a Müller sprayer (11 sections, DDOP has DDI 160 only under element 2):

Subscribed (OnChange) to DDI 160 (Section Control State) for element 2
Subscribed (OnChange) to DDI 160 (Section Control State) for element 17
Subscribed (OnChange) to DDI 160 (Section Control State) for element 18
...
Sending set value to client 135 for DDI Section Control State element 18 with value Auto/on
Client 135 value command for element 2 DDI Section Control State with value Manual/off OK.

Elements 17 and 18 are plain sections that don't list the DDI 160 object. The TC wrote "automatic" to element 18, the sprayer stayed in manual on element 2, and it ignored every setpoint. Whether it happens depends on heap contents: the same sprayer mapped correctly with another build.

Change

One helper, get_child_object_ids(DeviceElementObject &), copies the IDs through get_number_child_objects() / get_child_object_id(i). AgIsoStack's own code uses those (DDOP helpers, TC client, ISOXML export). All four loops now use the helper.

Notes

• The DataSpan bug itself should be fixed upstream in AgIsoStack. This PR doesn't depend on that fix and stays correct after it.
• The helper's comment explains why not to go back to the span.

Testing

• Release build passes on Windows (MSVC).
• Decoded the Müller DDOP with AgIsoStack's deserializer, in the same 4 chunks as the upload: the indexed accessor returns exactly the listed children, and DDI 160/161/290 map to element 2 only.
• To be confirmed on the bench: the sprayer reports Section Control State = Auto/on on element 2 and follows setpoints.

🤖 Generated with Claude Code

[GwnDaan]

@gunicsba Let's fix the issue in agisostack itself instead :)

[gunicsba]

@GwnDaan You want to fix it or should I ask Claude to do so?

[gunicsba]

@GwnDaan Fix proposed upstream in Open-Agriculture/AgIsoStack-plus-plus#783: DataSpan now indexes in elements, not bytes. It includes tests, and CI is green. Once it's merged I'll turn this PR into a bump of the AgIsoStack GIT_TAG and drop the workaround.

[gunicsba]

Closing in favour of the upstream fix, Open-Agriculture/AgIsoStack-plus-plus#783. #110 tracks bumping AgIsoStack once it's merged.