I was trying to build one of the ArduinoBLE examples (scan), using my updated core code for UNOR4 Wifi,
and: HCIVirtualTransport.cpp failed to compile as RingBufferN was not defined... This was with my current version I removed the usage of it from the Serial code.
Fix for this one is to explicitly include it in the file:
i.e. to add the line: #include "api/RingBuffer.h"
While I was looking at how the RingBuffer code was used in this library, I believe that the usage in
the HCICardioTransport class may be unsafe.
That is the buf.read_char is called from its main read method.
int HCICordioTransportClass::read()
{
return _rxBuf.read_char();
}
And the store_char is called in the method HandleRxData which is called by
the method onDataReceived, which appears to be called by an event.
The problem is there is a timing window in the current code, Where if the callback happens while in the read_char, right after it reads in the current value of _numElems and before it stores the updated count back out, the callback happens, which updates the count. When it returns, the storing out of the updated count in the read_char will wipeout the count changes that happened in the store_char call(s) - could be be multiple bytes that are added during the callback.
I was trying to build one of the ArduinoBLE examples (scan), using my updated core code for UNOR4 Wifi, and: HCIVirtualTransport.cpp failed to compile as RingBufferN was not defined... This was with my current version I removed the usage of it from the Serial code. Fix for this one is to explicitly include it in the file: i.e. to add the line: #include "api/RingBuffer.h"
While I was looking at how the RingBuffer code was used in this library, I believe that the usage in the HCICardioTransport class may be unsafe.
That is the buf.read_char is called from its main read method.
And the store_char is called in the method HandleRxData which is called by the method onDataReceived, which appears to be called by an event.
As I mentioned in the issue: https://github.com/arduino/ArduinoCore-API/issues/195 Both the read_char and store_char both modify a member variable: _numElems To either increment or decrement it.
The problem is there is a timing window in the current code, Where if the callback happens while in the read_char, right after it reads in the current value of _numElems and before it stores the updated count back out, the callback happens, which updates the count. When it returns, the storing out of the updated count in the read_char will wipeout the count changes that happened in the store_char call(s) - could be be multiple bytes that are added during the callback.
Hope this was clear.