meshcore-bot/BUGS.md

# BUGS

Known defects and outstanding issues in meshcore-bot. Fixes that have shipped
are listed in [CHANGELOG.md](CHANGELOG.md); the archive at the bottom of this
file is kept only for historical reference.

---

## Outstanding known issues

As of **v1.1.0**.

### High priority

_None outstanding._

### Medium priority

| ID | Module | Description | Workaround |
|----|--------|-------------|------------|
| BUG-004 | `message_handler` | RF data correlation (SNR/RSSI) can miss messages if the RF log event arrives more than `rf_data_timeout` (default 15 s) after the message | Increase `rf_data_timeout` in `[Bot]` |
| BUG-005 | `scheduler` | Memory headroom is tight on 512 MB boards such as the Raspberry Pi Zero 2 W when the bot and web viewer run together. v1.0.0 reduced the load substantially — the dashboard is served from a background snapshot rather than recomputing ~50 aggregate queries per request, and retention now prunes on schedule — but the ceiling is still close on the smallest boards | Disable the web viewer (`[Web_Viewer] enabled = false`) or tune the mesh graph (`graph_startup_load_days = 7`) |
| BUG-006 | `feed_manager` | Stale rows in `feed_message_queue` from an old install can cause repeated queue-processing errors after a database migration | Clear the pending queue: `DELETE FROM feed_message_queue WHERE sent_at IS NULL` |
| BUG-008 | `telegram_bridge_service` | Telegram `message_thread_id` (forum/topic support) is not implemented — messages go to the main group channel only | None; tracked in [TODO.md](TODO.md) with the two-way Telegram bridge work |

### Low priority / by design

| ID | Module | Description | Notes |
|----|--------|-------------|-------|
| BUG-011 | `repeater_manager` | MeshCore devices hard-limit contacts to 300; the auto-purge threshold is 280, so purging 20 at a time may not keep up on very busy meshes | Tune `auto_purge_threshold` and ensure `auto_manage_contacts` is enabled |
| BUG-014 | `packet_capture_service` | Packet hash calculation silently falls back to a default value on failure (`packet_capture_service.py:1347`) | Low impact; affects deduplication accuracy only |
| BUG-026 | `message_handler` | Keyword-dispatched help and command responses are sent as a single message with no auto-chunking, so long responses may be truncated by transport limits rather than split into extra multipart messages | Design choice. Commands can call `send_response_chunked()` explicitly if they want multi-part replies |

### Closed as won't fix

| ID | Module | Description |
|----|--------|-------------|
| BUG-007 | `discord_bridge_service` | Discord webhook rate limiting is expected behavior. Keep bridged channels low-traffic or rate-limit upstream |

---

## Reporting new bugs

Open an issue at the [project repository](https://github.com/agessaman/meshcore-bot/issues)
and include:

- Bot version (`git describe --tags`)
- The relevant section of `config.ini`, with keys and tokens redacted
- Log output (`logs/meshcore_bot.log`) from around the time of the issue
- Steps to reproduce

**Security issues should not be filed publicly** — see [SECURITY.md](SECURITY.md)
for private reporting.

---

## Archive: fixed bugs

Historical record only. This section is not maintained going forward — see
[CHANGELOG.md](CHANGELOG.md) for what shipped in each release.

Note: the `26d18c1` and `ab72be9` references below predate a history rewrite and
no longer resolve to commits in this repository. The descriptions are still
accurate; only the hashes are stale.

### v0.9.0 (2026-04-17)

| Reference | Summary |
|-----------|---------|
| Issue #83 | `KeyError('msg_hash')` asyncio parser spam fixed upstream by `meshcore >= 2.3.6` (new `meshcore_parser.py` guards with `'msg_hash' in l`). |
| Issue #91 | Bot `!version` command added and version string shown in the web-viewer footer (`883b67d`, `fbf3995`). |
| Issue #126 | `can't convert negative int to unsigned` on flood contacts fixed upstream by `meshcore >= 2.3.6` (out_path_len `-1` → `255` before packing). Defensive wire-field rebuild in `_ensure_contact_meshcore_path_encoding` (`ba52c3b`). |
| Issue #161 | Shipped config templates lowered `max_response_hops` default from 10 → 7 to reduce flood-response amplification. |
| Tests regression | Rewrote `test_subscribe_packets_emits_status_ack` / `test_subscribe_messages_emits_status_ack` to match the silent-UX contract introduced in `1ee84f2`. |
| Python version | `requires-python` raised to `>= 3.10` in `pyproject.toml`; ruff target-version `py310`; CI matrix adds `3.13`. |
| `e0eae09` (PR #123) | Fixed CI failures introduced by v0.9.0 push: (1) ruff — fixed import order, `Dict`→`dict`, and unused variable in `discord_bridge_service.py`; (2) mypy — added `types-requests` stub package to test deps, added `ignore_errors = true` per-module overrides for 19 not-yet-annotated modules; (3) ShellCheck SC2034 (unused vars) in `install-service.sh` / `uninstall-service.sh`, SC2155 (declare+assign) in `install-service.sh` / `restart_viewer.sh`, SC2010 (`ls\|grep`) replaced with glob loops in `docker-setup.sh`, SC2320 (`$?` capture after heredoc) in `docker-setup.sh` |
| `92c5910` (PR #122) | Removed Python 3.9 from CI test matrix — `meshcore >=2.2.31` requires Python >=3.10 and is not installable on 3.9 |
| `164dbae` | Refactored command aliases from global `[Aliases]` config section to per-command `aliases =` key in each command's own config section; `BaseCommand._load_aliases_from_config()` reads and injects aliases at startup; `CommandManager.load_aliases()` and `_apply_aliases()` removed |
| `f971e97` | Fixed pre-existing test failure in `test_discord_bridge_multi_webhooks.py`: `ConfigParser` lowercases all config keys so `bridge.Public` is stored as `"public"` — test assertions updated to match actual lowercase key behaviour; runtime matching was already case-insensitive |
| `26d18c1` | Fixed BUG-029 (third pass): Realtime monitor panels stuck at "Connecting…" — root cause was `<script type="module">` in `realtime.html` creating a second Socket.IO manager that raced with `base.html`'s `forceNew: true` manager; the module-socket's `connect` event never fired. Also: `ping_timeout=5` (5 s) was too short for subscribe handlers that replay DB history; `subscribed_messages` key was missing from `connected_clients` initial dict. Fixed: changed `<script type="module">` to regular `<script>` with dynamic `import()` for the decoder; removed `forceNew: true` from `base.html` so both pages share one Socket.IO manager; raised `ping_timeout` 5→20 s; added `subscribed_messages: False` to client tracking dict. |
| `26d18c1` | Fixed BUG-029 (second pass): `config_base` was a local variable — stored as `self._config_base` instance attribute; removed dead `_get_db_path()` method that still used `self.bot_root`; fixed `subscribe_logs` and `_start_log_tailing` to resolve log file path via `self._config_base` instead of `self.bot_root`; fixed misleading hardcoded "Connected"/"Active" status badges in `realtime.html` (now start as "Connecting…" and update dynamically on actual SocketIO connect). |
| `26d18c1` | Fixed BUG-029 (first pass): `app.py` resolved `db_path` relative to the code root (2 dirs above `app.py`) instead of relative to the config file's parent directory, causing the web viewer and bot to open different database files. Fixed: `config_base = Path(config_path).parent.resolve()` used as base for `resolve_path()`; also elevated subscribe-handler replay errors from DEBUG to WARNING and added INFO log of resolved db_path on startup. 4 new tests in `TestDbPathResolutionFromConfigDir`. |
| `26d18c1` | Fixed BUG-027: `test_weekly_on_wrong_day_does_not_run` was patching `get_current_time` with real `now` (Monday) instead of `fake_now` (mocked Tuesday) — test always passed on non-Monday but failed on Mondays; fixed in `tests/test_scheduler_logic.py:430` |
| `26d18c1` | Fixed BUG-025: `send_channel_message` did not retry on `no_event_received` — added retry loop in `command_manager.py` (max 2 retries, 2s delay); `_is_no_event_received()` helper; 5 tests |
| `ab72be9` | Fixed BUG-024: DB backup scheduler fired every second — `last_db_backup_run` now updated after each call; 2-min fire window prevents triggering on late startup; last-run seeded from DB on restart |
| `ab72be9` | Fixed BUG-023: Realtime monitoring command stream blank on load — added 50-row history replay to `subscribe_commands` handler; fixed `last_timestamp = 0` → `time.time() - 300` in polling thread |
| `ab72be9` | Fixed BUG-022: `IndexError` from meshcore parser silently discarded in asyncio task — installed custom loop exception handler in `core.py:start()` to suppress at DEBUG level |
| `ab72be9` | Fixed BUG-020: Config TUI `[Scheduled_Messages]` keys showed `?` marker — dynamic sections (no fixed example keys) now suppress unknown-key marker |
| `ab72be9` | Fixed BUG-021: Config TUI had no way to edit the time portion of a scheduled message — added `r` (rename key), `a` (add key+value), `d`/Delete (delete key with confirmation) bindings in keys pane |
| `ab72be9` | Fixed BUG-019: `addPacketEntry` in `realtime.html` crashed with TypeError when `data.path` is a string — guarded with `Array.isArray()` |
| `ab72be9` | Fixed BUG-018: Real-time monitoring SocketIO connections dropped due to Werkzeug 3.1 WebSocket teardown assertion (see BUG-012 fix); DB polling and subscription replay confirmed correct |
| `ab72be9` | Fixed BUG-017: `disconnect_radio()` now uses `asyncio.wait_for(..., timeout=10)` — no longer hangs indefinitely |
| `ab72be9` | Fixed BUG-016: `reboot_radio()` now sends `meshcore.commands.reboot()` firmware command before disconnecting/reconnecting |
| `ab72be9` | Fixed BUG-015: scheduler thread blocked on `future.result(timeout=X)` causing `TimeoutError` spam and stalling the loop — replaced all four blocking waits with `add_done_callback` (fire-and-forget) in `run_scheduler` |
| `ab72be9` | Fixed BUG-001: web viewer now supports optional password authentication via `web_viewer_password` in `[Web_Viewer]` config |
| `ab72be9` | Fixed BUG-002: `db_manager` now runs ALTER TABLE migrations for `channel_operations` (`result_data`, `processed_at`) and `feed_message_queue` (`item_id`, `item_title`, `priority`) on startup |
| `ab72be9` | Fixed BUG-003: geocoding rate-limit skip in `repeater_manager` now logs at INFO level instead of DEBUG so it is visible in production logs |
| `1264f49` | Fixed repeater manager auto-purge ignoring `auto_manage_contacts` config — purge ran unconditionally regardless of setting (issue #50) |
| `1264f49` | Fixed web viewer responses returning stale or incorrect repeater data |
| `5c8ee35` | Fixed timezone handling in `format_elapsed_display` — elapsed times displayed incorrectly in non-UTC timezones (issue #75) |
| `1cc41bc` | Fixed repeater usage and web viewer response formatting (PR #67) |
| `1474174` | Fixed `TraceCommand` path truncation — return paths were being cut short |
| `1a576e8` | Fixed reversed path nodes in `TraceCommand` — trace direction was backwards |
| `5a96dec` | Fixed incorrect hop labeling logic in `TraceCommand` |
| `2178a80` | Fixed log spam during shutdown — cleanup methods logged errors after streams were closed |
| `e9f17ec` | Fixed incomplete shutdown — scheduler thread and meshcore disconnect were not always joined cleanly |
| `217d2a4` | Fixed database connection handling across modules — connections were not always properly closed |
| `d084c6b` | Fixed `PrefixCommand` not supporting multi-byte hex prefix lookups |
| `6c81513` | Fixed `MeshGraph` edge promotion logic — edges were not promoted correctly under some conditions |
| `36a8a67` | Fixed prefix handling incompatibility when transitioning from 1-byte to 2-byte prefixes |
| `0c060a5` | Fixed chunked message sending race with rate limiter — second chunk could be blocked |
| `58deb12` | Fixed `RepeaterManager` ignoring `auto_manage_contacts = false` |
| `8186648` | Fixed BUG-028: `decode_meshcore_packet()` no longer throws `UnboundLocalError` when `bytes.fromhex()` fails (invalid hex input now cleanly returns `None`) |