reset procedure fails on esp32s3

[jepler]

This may be a core bug but I'm starting here...

On a prototype board with MAX17048, I was consistently failing to initialize the device:

>>> s = adafruit_max1704x.MAX17048(b)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "adafruit_max1704x.py", line 115, in __init__
  File "adafruit_max1704x.py", line 130, in reset
  File "adafruit_register/i2c_bit.py", line 64, in __set__
OSError: [Errno 19] No such device

reset() deliberately causes a transaction that will be NAK'd. However, on Adafruit CircuitPython 8.0.0-beta.4-25-g195ad4e479 on 2022-11-15; Adafruit Feather ESP32-S3 Reverse TFT with ESP32S3, the next I2C transaction after the NAK would also fail.

The following change seems to allow the module to consistently be initialized:

diff --git a/adafruit_max1704x.py b/adafruit_max1704x.py
index cb4d700..ae0eccd 100644
--- a/adafruit_max1704x.py
+++ b/adafruit_max1704x.py
@@ -120,11 +120,18 @@ class MAX17048:
         try:
             self._cmd = 0x5400
         except OSError:
+            print("nak'd command")
             # aha! we NACKed, which is CORRECT!
             pass
         else:
             raise RuntimeError("Reset did not succeed")
-        self.reset_alert = False  # clean up RI alert
+        for _ in range(3):
+            print("trying to reset alert")
+            try:
+                self.reset_alert = False  # clean up RI alert
+                return
+            except OSError as e:
+                print(f"tried to reset alert: {repr(e)}")
 
     @property
     def cell_voltage(self) -> float:

The debug prints that remain say that the first reset_alert is also failed / NAK'd but the 2nd succeeds

s = adafruit_max1704x.MAX17048(b); time.sleep(.1); s.cell_voltage
nak'd command
trying to reset alert
tried to reset alert: OSError(19,)
trying to reset alert
4.2175

I did not scope the I2C bus to see whether the first attempted transaction to reset_alert actually occurs or not.

[jepler]

@ladyada @dhalbert interesting clue about i2c problems on esp32s3 emerged while working with this sensor on the feather esp32s3 reverse tft prototype board

[ladyada]

could be its just faster because it doesnt have the 1ms wierdo delay? wanna try adding a lil sleep(0.005)?

[jepler]

A sleep (even a 1-second sleep!) isn't enough to recover. I think we need the "try twice" fix here, unless we find a fix for the core.

[jepler]

Closed by #10