FreeDMR 1.x Changelog

Test Harnesses

Added an in-process deterministic packet harness for bridge_master.py routing, state, expiry and packet rewrite checks without UDP.
Added a black-box UDP harness that starts FreeDMR with generated test configs, emulates HBP clients and FBP/OpenBridge peers, captures outbound UDP, supports venv bootstrap, and can model packet loss, duplicates and reordering.
Added synthetic and recorded packet fixture coverage for routing, slot rewrite, byte preservation, malformed packets, cadence and link impairment.

Hardened config parsing for booleans, alias stale time, missing session options and invalid numeric fields.
Added DIAL_A_TG to disable private-call dial-a-TG control.
Added DYNAMIC_TG_ROUTING to disable automatic creation of unknown conventional TG bridges.
Deprecated DEFAULT_REFLECTOR as the system default dial-a-TG setting, while keeping it as a TS2 compatibility alias.
Added canonical per-slot defaults: DEFAULT_DIAL_TS1 and DEFAULT_DIAL_TS2.
Kept legacy OPTIONS aliases DIAL, StartRef and DEFAULT_REFLECTOR mapped to TS2; explicit DEFAULT_DIAL_TS2 takes precedence.
Added validation/logging so invalid default dial values do not create bridge state and normalize to no default for the active runtime session.

Made dial-a-TG private-call control slot-local: TS1 controls TS1, TS2 controls TS2. TS1 no longer retunes or disconnects TS2.
Preserved voice prompts on TG9 TS2.
Preserved TG9 as the RF-visible dial-a-TG talkgroup for both slots.
Rejected reserved/control targets consistently for live dial-a-TG and default startup/session configuration.
Preserved the current FreeDMR dial-a-TG policy cap of 999999.
Ensured FBP route targets created for dial-a-TG remain active across local retunes/disconnects, in line with the mesh "everything everywhere" model.

Preserved DMR data forwarding support.
Kept DATA-GATEWAY behavior for protocol-v1 SMS/GPS style handling.
Reported group-addressed data as data/control, not as voice lifecycle.
Suppressed false GROUP VOICE timeout reports for data/control packets.
Preserved data-sync/control payload bytes across HBP and FBP forwarding.
Kept SUB_MAP last-known-location behavior for unit data routed toward HBP.
Preserved FBP metadata such as source server, source repeater, BER, RSSI and hops according to protocol version.

Preserved real inbound LC where available and used explicit synthetic LC only as fallback.
Switched normal synthetic group voice LC service options to 0x00; retained HBLink 0x20 only as an explicit legacy compatibility constant.
Reworked embedded LC handling so same-TG forwarding preserves embedded LC payloads where possible, while TG-mapped forwarding regenerates routing LC.
Added in-call Talker Alias and GPS embedded-LC logging without changing routing or packet mutation behavior.
Added generated prompt lifecycle handling so real RF voice can interrupt a prompt instead of being blocked as busy.
Fixed private dial-a-TG/AMI timeout reporting so private control calls do not emit unmatched group voice lifecycle events.
Made HBP and FBP voice sequence handling modulo-256 with explicit duplicate, loss and stale/out-of-order treatment.
Ensured voice terminators mark streams finished even when reporting is disabled, preventing late same-stream packets from reopening ended streams.

Added malformed/truncated DMRD and DMRE guards before fixed-offset parsing.
Corrected source-quench matching so BCSQ uses the TG namespace visible to the peer being quenched, including dial-a-TG reflector TGs.
Made STUN/BCST handling consistent as a broad FBP traffic gate.
Preserved protocol-version-sensitive FBP/OBP metadata layout.
Added tests for FBP keepalive gating, wrong network ID, bad hashes, stale timestamps, max-hop handling, v4 characterization and v1 refusal on v5 links.

Kept the legacy report socket opcode model unchanged.
Kept bridge event CSV field order unchanged.
Kept DEFAULT_REFLECTOR in runtime config as the effective TS2 default for compatibility with existing config/API/report consumers.
Kept prompt/ident generated audio visible as TG9 TS2.
The latest per-slot default-dial changes do not introduce new report event names or new report event fields.
Expected dashboard impact is low if the dashboard reads event fields and bridge entries by their existing keys.
Compatibility risk: BRIDGE_SND pickled bridge state may now include active TS1 #reflector entries. A dashboard that assumes every #reflector entry is TS2-only may need an update; a dashboard that already respects TS, TGID and ACTIVE should continue to parse it.
TS1 dial-a-TG activity may now appear as RF-visible TG9 on slot 1, which is intentional new behavior.

Added freedmr_dmr_codec.py for locally tested DMR LC, embedded LC, slot type, BPTC, Hamming, Golay and RS parity helper behavior.
Moved runtime LC generation and byte/int helper usage away from older dmrutils3 functions where covered.
Added standalone codec tests using fixed fixtures and MMDVMHost-style behavior.
Added focused utility tests for ID/byte helpers and alias lookup.

Replaced the Spyne-based API path with a small bounded HTTP/JSON control API.
Kept API operations as small in-memory control-plane actions.
Added API tests for request size limits, key validation, JSON responses, option storage and reset/kill behavior.
Tidied auxiliary tests for report receiver flags, SQL report insertion, AMI factory state and proxy environment booleans.

Backported only directly relevant, already-supported fixes from bridge_master.py to bridge.py.
Kept bridge.py focused on its existing conference-bridge role; did not add FreeDMR master-only features such as dial-a-TG.

Added and updated harness, testing, API and architecture documentation.
Added FreeDMR 2 design/ADR documents separately, without changing current 1.x runtime behavior.
Maintained docs/codex-notes.md as the engineering notebook for findings, assumptions, protocol-sensitive areas, invariants and unresolved questions.

Current non-UDP test discovery passes.
Focused UDP black-box tests for TS1 dial-a-TG and disabled dynamic TG routing pass when local UDP sockets are permitted.
Live RF validation is still required before treating protocol-visible behavior changes as release-ready.