risco_patch

A tiny Home Assistant custom component that monkey-patches pyrisco at runtime to fix three known bugs in the official Risco Local integration.

The official integration ships pyrisco 0.6.8, where these bugs are unresolved as of HA 2026.4.x.

Status (2026-05-04). The 3 patches eliminate the integration-side crashes (UTF-8, FAILED_UNLOAD, 0x17 ValueError). They do NOT eliminate the panel "supervision lost" beep that some users (incl. me) are experiencing every ~20 minutes since a recent regression — see the open question section.

---

What the patch fixes (3 code patches)

1. UnicodeDecodeError: 'utf-8' codec can't decode byte 0xa7

pyrisco hardcodes UTF-8 when decoding panel responses. Italian (and other latin-1 / cp1252) panels routinely send bytes in 0x80–0xFF. Each invalid byte raises UnicodeDecodeError, killing the listener task.

Patch: default encoding="latin-1" on RiscoSocket.__init__ (latin-1 maps every byte 0x00-0xFF, never raises). Plus errors="replace" on the actual decode call as a belt-and-braces safety net.

2. FAILED_UNLOAD after ConnectionResetError

async_unload_entry calls local_data.system.disconnect() without a try/except. When the panel has already RST'd the TCP connection, _writer.wait_closed() raises ConnectionResetError. HA marks the entry as FAILED_UNLOAD and refuses any reload — only a full HA restart unblocks it.

Patch: wrap RiscoSocket.disconnect in try/except (ConnectionResetError, OSError, BrokenPipeError) and call _close() for cleanup. Same intent as upstream PR home-assistant/core#165924 (open, approved by @OnFreund, awaiting merge as of 2026-05).

3. ValueError: too many values to unpack (expected 2, got 3) in RiscoCrypt.decode

decode() does command, crc = decrypted.split('\x17') assuming exactly one 0x17 (ETB) byte in the message. If the payload contains another 0x17, the split returns 3 parts and crashes the listener.

Patch: use rfind('\x17') so the last 0x17 is treated as the separator; tolerate empty / malformed messages by returning a "bad CRC" tuple instead of raising.

---

Open question — recent regression, please help

In my own setup (HA OS 2026.4.4, pyrisco 0.6.8, Risco LightSYS Plus IT firmware), with the 3 patches above applied and the integration otherwise stable, the panel still emits a "supervision lost" beep every ~20 minutes, triggered by an unsolicited TCP RST from the panel.

Crucially: this was NOT happening a few weeks ago with the same panel and the same network. The integration was rock-solid for months. Something changed recently — and it's not pyrisco itself (0.6.7 → 0.6.8 only changed cloud User-Agent, the local socket code is byte-identical).

If you have any of the following info, please open an issue / reply on the HA forum thread — it would help nail the root cause:

• Same panel model + same firmware: are you also seeing the ~20-min RST/beep cycle now, even though it was fine weeks ago?
• Did your panel receive a firmware OTA update recently?
• Any HA core release in the last few weeks that changed the Risco integration's poll interval / keep-alive / connection lifecycle?
• pyrisco DEBUG logs showing the last few RX/TX commands before the panel sends the RST.

Things I've already ruled out: scan_interval (tried 30s, 60s, 600s), keepalive interval (5s vs 60s) — the RST timing is identical regardless. So it's not poll-induced; the panel itself is closing the connection.

Workarounds that mask but don't fix this (and that I'm explicitly NOT shipping in this repo):

• A 15-min "preemptive reconnect" automation that reloads the config entry before the panel times out. Hides the beep but causes 1-3 min of setup_in_progress per cycle and assumes the timeout is fixed (it might not be).
• Bumping scan_interval to a high value. Doesn't actually change the RST timing.

If/when the real cause is found, I'll update this repo and the forum post.

---

Install

Option A — Manual

1. Copy custom_components/risco_patch/ into <config>/custom_components/. Final layout:

   <config>/
   ├── configuration.yaml
   └── custom_components/
       └── risco_patch/
           ├── __init__.py
           └── manifest.json
   

2. Add one line to configuration.yaml:

   risco_patch:

3. Settings → System → Restart Home Assistant.

4. Verify in Settings → System → Logs (or home-assistant.log):

   [risco_patch] RiscoSocket encoding default -> latin-1
   [risco_patch] RiscoSocket.disconnect hardened
   [risco_patch] RiscoCrypt.decode patched (rsplit + errors=replace)
   

Option B — HACS custom repository

1. HACS → ⋮ → Custom repositories → add this repo URL, category Integration.
2. Search "Risco Patch", install.
3. Add risco_patch: to configuration.yaml.
4. Restart HA.

---

How to verify it actually worked

Before	After (with this patch)
Repeated UnicodeDecodeError: 'utf-8' codec can't decode byte 0xa7… in the log	Gone
Risco config entry occasionally lands in failed_unload, requires HA restart	Stable loaded state
ValueError: too many values to unpack in RiscoCrypt.decode	Gone
Panel beep every ~20 min from supervision lost	Still present — root cause not yet identified, see open question

---

How to remove

1. Remove risco_patch: from configuration.yaml.
2. Delete <config>/custom_components/risco_patch/.
3. Restart HA.

The patch leaves no persistent state.

---

Compatibility & caveats

• Tested on: HA Core 2026.4.4 (Home Assistant Green), pyrisco 0.6.8, panel Risco LightSYS Plus (Italy).
• Patches are written defensively and idempotent — they check a marker attribute before reapplying.
• The latin-1 default is only applied if the caller doesn't specify encoding. If a future HA version starts passing encoding="utf-8" explicitly, this patch becomes a no-op for Patch 1.
• If your panel firmware speaks pure ASCII (most US installations), the encoding patch is harmless but unnecessary.

---

Why a custom_component instead of a fork of pyrisco?

A custom_component:

• Survives pyrisco package updates (you don't need to keep a vendored copy).
• Is opt-in (set via configuration.yaml line).
• Disappears cleanly when the upstream fix lands.
• Doesn't require touching the HA core image (which is read-only on HA OS).

---

Links

• Upstream resilience PR: home-assistant/core#165924
• pyrisco repo: https://github.com/OnFreund/pyrisco
• HA Risco integration docs: https://www.home-assistant.io/integrations/risco
• HA forum discussion: https://community.home-assistant.io/t/risco-local-custom-component-to-fix-utf-8-failed-unload-0x17-split-bugs-and-an-open-question-on-a-recent-regression/1009293

---

License

MIT. Use at your own risk.