ESP32 I2C writes work but not reads?

[NiHaoMike]

I'm working on a project that involves an ESP32 communicating with various I2C slave devices, such as a MCP23017.

This works as expected:

Code: [Select]

void MCP_writeA(uint8_t dat) {
  Wire.beginTransmission(0x20);
  Wire.write(0x12);
  Wire.write(dat);
  Wire.endTransmission();
}
This always returns 0 regardless of what the slave actually sent:

Code: [Select]

uint8_t MCP_readA() {
  Wire.beginTransmission(0x20);
  Wire.write(0x12);
  Wire.endTransmission(false);
  Wire.requestFrom(0x20, 1);
  return Wire.read();
}
This, for reading the cell voltages from the BMS, also doesn't work:

Code: [Select]

void BMS_Vcell(uint16_t ret[]) {
  Wire.beginTransmission(0x19);
  Wire.write(0x00);
  Wire.endTransmission(false);
  Wire.requestFrom(0x19, 8);
  ret[0] = (Wire.read() << 8) | Wire.read();
  ret[1] = (Wire.read() << 8) | Wire.read();
  ret[2] = (Wire.read() << 8) | Wire.read();
  ret[3] = (Wire.read() << 8) | Wire.read();
}
In all cases, I have verified that the traffic on the bus is as expected with a logic analyzer. (As in the slaves are in fact returning data.) I'm using the stable (1.0.0) arduino-esp32 core on the Arduino 1.8.7 IDE. I'm simply calling Wire.begin() to initialize the I2C.

[hamster_nz]

I have sometimes had these sorts of issues with a multi-function pin that isn't configured with the desired function. See if you use a different I/O pin and get similar results.

... humm... just had a look on the web while waiting for something to happen, and it looks like it should work.

Have you looked at what Wire.available() says?

[tsman]

There are several I2C bugs in arduino-esp32 core 1.0.0. They're only fixed in the development branch so try 1.0.1RC1.

[NiHaoMike]

There are several I2C bugs in arduino-esp32 core 1.0.0. They're only fixed in the development branch so try 1.0.0RC1.

That was it. Over 2 years since the ESP32 was released to public and something as basic as I2C is still broken on the stable branch? Just imagine how confusing it would be without a logic analyzer to get the ground truth of what's going on...

[tsman]

That was it. Over 2 years since the ESP32 was released to public and something as basic as I2C is still broken on the stable branch? Just imagine how confusing it would be without a logic analyzer to get the ground truth of what's going on...

Oops. I made a typo on the version since it should be 1.0.1RC1 but you worked that out.

ESP32 Arduino core 1.0.0 only came out a few months ago but that isn't any excuse for this mess. This I2C bug was known about months before 1.0.0 was even released and somebody had already made a fixed version of the library.