NACK last byte when read

[bbx10]

The TCS34725 RGB color sensor works reliably with this change. See #535 for details.

[me-no-dev]

have you tested this with other i2C devices?

[bbx10]

Yes, a 9DOF IMU board which was not working now works with this change and the clock stretch increase (#553). This change might help with the other IMU board (#526).

I will test 4 other i2c devices with this change and #553 tomorrow.

I am using the following for the i2c protocol reference:
http://www.nxp.com/documents/user_manual/UM10204.pdf.

Section 3.1.6 talks about 5 conditions for generating NACK. The last one says

1. A master-receiver must signal the end of the transfer to the slave transmitter.

This is why the Uno/AVR generates NACK after the last byte read.

[igrr]

We used to have this correct long long time ago :)

Arduino/hardware/esp8266com/esp8266/cores/esp8266/i2c.cpp

Line 164 in 7e40b9f

if(data == end - 1)

I will also test a few more devices myself with this change and #553.

[bbx10]

We used to have this correct long long time ago :)

Thanks, this helps assure me this is a good change. The change affects every i2c read so it has the potential for massive breakage. :(

The 9DOF IMU ran for 3 hours and was still reporting correct values. The i2c chips on board are the L3GD20H and LSM303.

The following other i2c devices work. Only a few minutes testing done per device.
BMP180, HTU21DF, ISL29125, TSL2591, TCS34725

All but the TCS34725 worked even without this change. All testing was done with #553 as well as this change.

[igrr]

I have tested this change with MPR121, HMC6352, and one I2C EEPROM (forgot the part name).
All seem to be work fine.