RP2040 rotaryio fails with two encoders

[todbot]

CircuitPython version and board name

Adafruit CircuitPython 9.2.4 on 2025-01-29; Raspberry Pi Pico with rp2040

and:
Adafruit CircuitPython 9.2.4-5-g3236a0f200 on 2025-01-31; Raspberry Pi Pico with rp2040  
  ("absolute newest" as of noon PST 3 Feb 2025)
Adafruit CircuitPython 9.2.3 on 2025-01-17; Raspberry Pi Pico with rp2040
Adafruit CircuitPython 9.2.2 on 2025-01-09; Raspberry Pi Pico with rp2040

Code/REPL

import time, board, rotaryio

encoder1 = rotaryio.IncrementalEncoder(board.GP19, board.GP18)
encoder2 = rotaryio.IncrementalEncoder(board.GP4, board.GP5)

while True:
    print(encoder1.position, encoder2.position)
    time.sleep(0.1)

Behavior

Code hangs on creation of encoder2, which is obviously not correct.

Code should repeatedly print position of both encoders.

Description

Works on 9.2.1.
Does not work on 9.2.2 and above.
Related to #9682?

Additional information

Redditor had his issue too: https://old.reddit.com/r/circuitpython/comments/1igqzyf/two_incrementalencoders_fail/

[todbot]

Update: It appears to get out of hanging if a change happens on encoder2. But then encoder1 does not work.

[jepler]

From 9.2.1 to 9.2.2 the big thing seems to be my PIO re-org to support the high pins on rp2350b (#9901). I know there were also changes to interrupts (or was it DMA? which this driver doesn't use-- #9980) but those seem to have occurred after 9.2.3.

[jepler]

(gdb) where
#0  _transfer (data_out=<optimized out>, out_len=0, 
    out_stride_in_bytes=0 '\000', swap_out=false, swap_in=false, 
    in_stride_in_bytes=1 '\001', in_len=536939816, 
    data_in=0x20080dd7 "\020l\344\n\020d\344\n\020\020\n\001  \r\001 d\344\n\020l\344\n\020\254\377", self=0x20010d28) at common-hal/rp2pio/StateMachine.c:984
#1  common_hal_rp2pio_statemachine_readinto (self=self@entry=0x20010d28, 
    data=data@entry=0x20080dd7 "\020l\344\n\020d\344\n\020\020\n\001  \r\001 d\344\n\020l\344\n\020\254\377", len=len@entry=1, 
    stride_in_bytes=stride_in_bytes@entry=1 '\001', swap=swap@entry=false)
    at common-hal/rp2pio/StateMachine.c:1031
#2  0x10042930 in common_hal_rotaryio_incrementalencoder_construct (
    self=self@entry=0x20010d20, pin_a=pin_a@entry=0x100ae46c <pin_GPIO43>, 
    pin_b=pin_b@entry=0x100ae464 <pin_GPIO44>)
    at common-hal/rotaryio/IncrementalEncoder.c:85

The hang occurs when constructing the 2nd encoder instance. However, if I modify the StateMachine implementation so that "use_existing_program" always returns false, the problem stops. Or at least it successfully constructs the StateMachine object in this case. I don't have any encoders hooked up.

[jepler]

I wrote some code so that I didn't need any encoder knobs for testing.

# Test simulated encoder (issue #10024) for metro rp2350
# Connect jumper wires from RX/TX/D2/D3 to A0/A1/A2/A3
# Each encoder should count.
import board, rotaryio
import digitalio

d = digitalio.DigitalInOut(board.D0)
d.switch_to_output(False)
e = digitalio.DigitalInOut(board.D1)
e.switch_to_output(False)
f = digitalio.DigitalInOut(board.D2)
f.switch_to_output(False)
g = digitalio.DigitalInOut(board.D3)
g.switch_to_output(False)

r1 = rotaryio.IncrementalEncoder(board.A0, board.A1, divisor=1)
r1.position = 0
r2 = rotaryio.IncrementalEncoder(board.A2, board.A3, divisor=1)
r2.position = 0

for _ in range(8):
    d.value = True
    print(f"{+d.value}{+e.value} {r1.position} {r2.position}")
    e.value = True
    print(f"{+d.value}{+e.value} {r1.position} {r2.position}")
    d.value = False
    print(f"{+d.value}{+e.value} {r1.position} {r2.position}")
    e.value = False
    print(f"{+d.value}{+e.value} {r1.position} {r2.position}")

for _ in range(8):
    f.value = True
    print(f"{+f.value}{+g.value} {r1.position} {r2.position}")
    g.value = True
    print(f"{+f.value}{+g.value} {r1.position} {r2.position}")
    f.value = False
    print(f"{+f.value}{+g.value} {r1.position} {r2.position}")
    g.value = False
    print(f"{+f.value}{+g.value} {r1.position} {r2.position}")

[todbot]

Verified that fix #10025 does make two encoders work on RP2040 Pico and RP2350 Pico2.

[jepler]

Thanks for testing, @todbot !