[rhorvath02]

Purpose

Build each fixed vehicle architecture once and reuse its FMU with different calibration before initialization.

Approach

Add the architecture catalog, configurable exporter and source/build manifests. Preserve independent calibration and derived dependencies, use fixed-capacity curves/maps with active counts, and reject missing or silently frozen numeric fields. Apply guarded OMC runtime corrections and include compiler reproducers.

Validation

187 export/runtime checks passed. A fresh rear-inboard full-vehicle FMU passed all seven drive/brake/re-launch tests. Co-Simulation is tested; Model Exchange and complete architecture-catalog acceptance remain unverified.

[raorjun]

The runtime FMU tests do not run the vehicle export path or the calibration-inventory guard. CI stays green if the guard or the real vehicle manifest breaks.

[raorjun]

Fixed in c367cee. The new Tests/test_fmu_vehicle_calibration.py runs under BOBLIB_TEST_FMU_RUNTIME=1. It calls the real exporter.build for the default architecture, with no stubs. It checks that the inventory is not empty, and that driverMass, baseMass and both springTables are editable while frame mass is derived. With fmpy, +50 kg on driverMass raises the derived frame mass by exactly 50 kg and raises total Fz. It passed locally in 600 s. With validate_record_inventory forced to raise, it fails. The fmu CI job now runs it. Expect about 8-12 more minutes.