This fixes two issues that caused BLE connections to hang on macOS
when not using the --debug flag:
1. Race condition in BLEClient event loop initialization
- The event loop thread was started but asyncio operations were
submitted before the loop was actually running
- Added threading.Event synchronization to ensure the event loop
is running before any operations are submitted
- The ready signal is sent from within the loop via call_soon()
to guarantee the loop is truly active
2. CoreBluetooth callback delivery on macOS
- On macOS, CoreBluetooth requires occasional I/O operations for
callbacks to be properly delivered to the main thread
- Without --debug, no I/O was happening, causing callbacks to
never be processed and operations to hang indefinitely
- Added sys.stdout.flush() call before waiting for async results
to trigger the necessary I/O
The --debug flag masked these issues because:
- Debug logging introduces timing delays that let the event loop start
- Logger I/O triggers the necessary callback delivery mechanism
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
this function, xor_hash, and a variable for the default key (as bytes, I think, rather than the base64 version) really all belong in meshtastic.util rather than here.
There's multiple forms of hashing in firmware so this should be named to denote that, perhaps channel_hash. If we later want to add the frequency-slot-style hash, better if it's distinguished better from the start.
### Summary
- Added a new method `get_channels_with_hash()` to the `Node` class.
- This method returns a list of dictionaries, each containing the channel index, role, name, and a hash value derived from the channel name and PSK.
- The hash is calculated using the existing `generate_hash` utility, ensuring consistency with other parts of the codebase.
- This utility makes it easier to programmatically access channel metadata, including a unique hash, for scripting, debugging, or integration purposes.
### Motivation
- The protobuf `Channel` objects do not include a hash value by default.
- This addition provides a Pythonic, easy-to-use way to access channel info and a unique hash for each channel, which can be useful for diagnostics, scripting, or UI display.
### Example Usage
```python
channels = node.get_channels_with_hash()
for ch in channels:
print(f"Index {ch['index']}: {ch['role']} name='{ch['name']}' hash={ch['hash']}")
```
### Impact
- No breaking changes; existing APIs and CLI output remain unchanged.
- The new method is additive and can be used where needed for enhanced channel introspection.
The double flush() is not the root cause; the real issue is that code is trying to use the serial port after it has been closed.
The error occurs both in close() (during flush()) and later in _writeBytes() (during write()), indicating the port is closed or invalid at those times.
This ensures flush() is only called if the stream is open, and logs (but ignores) any exceptions during flush. This should prevent the "Bad file descriptor" error.
I see this error a lot on a rak unit, I dont know this is the way but .. you be the judge.
- Add T-Deck device definition with verified USB IDs (303a:1001)
- Add T-Deck to supported_devices list
- Uses t-deck firmware identifier and ESP32 device class
- Supports Linux (ttyACM), macOS (cu.usbmodem), and Windows (COM) ports
- Update .gitignore to exclude virtual environment files
T-Deck is now fully supported and automatically detected by the CLI.
github-actions