[puddly]

See #765 and NabuCasa/silabs-firmware-builder#236 for more context on the firmware side.

The address table actually contains the trust center address cache as a trailer so there's a small chance we can override an entry. Instead of randomly selecting entries, we round robin instead.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.58%. Comparing base (aea9248) to head (afcd5d5).

Additional details and impacted files

@@           Coverage Diff           @@
##              dev     #768   +/-   ##
=======================================
  Coverage   99.58%   99.58%           
=======================================
  Files          64       64           
  Lines        4331     4332    +1     
=======================================
+ Hits         4313     4314    +1     
  Misses         18       18           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

🟢 Approval recommended

The change is narrow, preserves configured table bounds, and includes tests for sequential allocation and wraparound.

0 open findings

What changed in this PR

Changes extended-timeout address-table allocation from random selection to bounded round-robin, complementing the related firmware work.

Changes:

• Adds a persistent cursor for sequential slot selection.
• Updates tests to verify sequential allocation and wraparound.

File	Description
tests/​test_ezsp_v4.py	Verifies sequential selection and wraparound.
tests/​test_ezsp_v14.py	Updates the expected initial slot.
bellows/​ezsp/​v4/​__init__.py	Replaces random selection with round-robin allocation.
bellows/​ezsp/​protocol.py	Initializes the address-table cursor.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🟢 Approval recommended

The change is narrow, preserves allocation bounds, and includes focused regression coverage with no unresolved findings.

No open issues 🎉

_{Review effort: Balanced}

[zigpy-review-bot]

Looks good to me. Replacing the random slot with a cursor is right for the host path: bellows is the only thing allocating application address table slots, and getConfigurationValue(CONFIG_ADDRESS_TABLE_SIZE) returns only the application slots, so the cursor walks free slots first and only then evicts round-robin (oldest first), which beats random eviction.

I checked out the head (afcd5d5) and ran tests/test_ezsp_v4.py, tests/test_ezsp_v5.py and tests/test_ezsp_v14.py: 286 passed. The new test_set_extended_timeout_table_wraps covers the wrap-around, and dropping the random patching makes the existing tests simpler.

Two non-blocking notes:

• The cursor lives on the protocol object and starts at 0 each time one is created, which is fine as long as a new protocol object always means a fresh NCP address table (the host path is stateless otherwise). If an EZSP reconnect ever keeps NCP state while recreating the protocol, the first writes would evict live entries again instead of using free slots.
• The cursor advances even if replaceAddressTableEntry returns a non-success status (its result is ignored, as before), so a failed write burns a slot position. Harmless in practice.

As the issue in #765 notes, the firmware-side % sli_zigbee_address_table_size pick in the XNCP combined send is separate and still needs its own change (NabuCasa/silabs-firmware-builder#236).