Validation and bring-up assertion tests

Test cases for validation and bring-up assertions (Validation and bring-up ass... (FEAT_0085)). Live under crates/ethercat-netcfg-codegen/tests/validation/.

Test Case: Overlapping slices are a build error unless allowed TEST_0844
status: open
verifies: REQ_0836

A fixture with two routings overlapping the same bit range in the same SubDevice and direction fails the build. Adding allow_overlap: true to the read-back channel makes the same fixture build. Confirms the default-error / opt-in-allow rule of Positional addressing; alia... (ADR_0093)’s sibling decision.
Test Case: Out-of-image, zero-length, dangling, and collision faults fail the build TEST_0845
status: open
verifies: REQ_0836, REQ_0835

A table-driven test feeds one fixture per fault — a slice past the ESI process-image size, bit_length: 0, a channel naming a non-existent device label, a channel-name collision, an override-induced address collision, an ESI hash mismatch against the lockfile, and a device carrying both an ESI reference and a contradicting inline offset — and asserts each fails the build with a distinct, named diagnostic.
Test Case: Unmapped process-image gaps warn but do not fail TEST_0846
status: open
verifies: REQ_0837

A fixture leaving an unmapped bit range in a device’s process image builds successfully and emits a non-fatal warning naming the gap.
Test Case: Generated bring-up assertions catch identity, alias, and WKC mismatch TEST_0847
status: open
verifies: REQ_0838

Driving the generated identity / alias table and expected_wkc against MockBusDriver, the test injects a wrong device identity at a position, a wrong station alias, and a working counter diverging from the expectation, and asserts each drives the connector to the Degraded / Down health path rather than proceeding to OP.
Test Case: SM-watchdog registers resolve and are emitted for output devices TEST_0860
status: open
verifies: REQ_0844
links incoming: REQ_0844

Resolution + arithmetic + codegen. An rx-carrying device with no override resolves to FTTI/2 (default 100 ms → 500 ticks of 100 µs at divider 2498); a per-device sm_watchdog_timeout_ms override and a custom ftti_ms resolve to their own quantized values; an input-only device resolves no watchdog. A value-for-value table pins the quantization (ceil(timeout_us / 100), clamp 1..=u16::MAX, 0 µs → 1 tick) against the connector’s documented SmWatchdog semantics, including the ceil-over-a-tick case. Codegen over a two-device fixture emits SubDeviceMap::new(..).with_sm_watchdog( SmWatchdog { divider, intervals }) for the output device and a bare SubDeviceMap::new(..) for the input-only device, asserted against the parsed AST.
Test Case: SM-watchdog bound and enable are validated at config time TEST_0861
status: open
verifies: REQ_0845
links incoming: REQ_0845

The Validate the SM-watchdog bo... (REQ_0845) matrix. PASS: ESI output SM with the watchdog trigger enabled and default FTTI/2; an override below the bound. FAIL: an override above FTTI/2; an ESI output SM declaring the watchdog trigger disabled; an inline output device with no sm_watchdog_enabled attestation; an inline output device attesting false. PASS: an inline output device attesting true. The inclusive boundary (override exactly at FTTI/2 on the tick grid) passes, proving the comparison is <= against the quantized value; an input-only device is skipped by both checks. Each failure asserts a distinct, named NetcfgError variant carrying the device label.
Test Case: No runtime parsing; generated config is static with no heap TEST_0848
status: open
verifies: REQ_0839

cargo tree for a consumer of the generated module lists no YAML parser (serde_yaml / serde_yml) in the runtime graph. A compile-time check confirms the generated PDO_MAP is a &'static binding usable in a const context, and a no-allocation test (under the bounded allocator) confirms consuming the generated tables performs no heap allocation.