`keypad.KeyMatrix()` not detecting any events if `row_pins` are more than nine pins

[rgbpcb]

CircuitPython version

Adafruit CircuitPython 8.0.0-alpha.1-31-gc1b00442a on 2022-07-02; stm32f411ce-blackpill with STM32F411CE

Edit: I'm using a slightly different board than the BlackPill, namely one without a low-speed crystal at pins C14 and C15.

Code/REPL

import board, keypad

key_matrix = keypad.KeyMatrix(
    row_pins=(board.A0, board.A1, board.A2, board.A5, board.A7, board.A8, board.B8, board.B9, board.C13, board.C14, board.C15, ),
    column_pins=(board.B14, board.B12, board.B10, board.B1, board.B0, board.A4, board.A3, ),
)

while True:
    event = key_matrix.events.get()
    if event:
        print(event)

Behavior

Nothing gets printed when keys are pressed.

Description

No response

Additional information

Removing a few pins from row_pins such that nine or fewer of them remain fixes the issue.

[dhalbert]

I tested a similar program on an Adafruit Feather STM32F405. I don't have a keyboard matrix with that many rows to try to try, so I wired up ten tactile pusbuttons, using 10 pins for the row and 1 column pin. I was not able to reproduce the problem

Did you try this without A10, C14, or C15 at all?

from board import *
import keypad

key_matrix = keypad.KeyMatrix(
    row_pins=(A1, A2, A3, A4, A5, SCK, MOSI, MISO, RX, TX),
    column_pins=(D12, ),
)

while True:
    event = key_matrix.events.get()
    if event:
        print(event)

[rgbpcb]

Did you try this without A10, C14, or C15 at all?

I'm not using pin A10. For pins C14 and C15, I'm using this bugfix (i.e. commenting out never_reset_pin_number() here) that gets them to work (except in the situations listed below). But you're right, leaving them out of row_pins and column_pins allows to detect some events (I haven't yet checked whether all possible events get detected).

The situations where pins C14 and C15 prevent keypad.KeyMatrix() from detecting any events (also events from other pins) are:

• The code mentioned above (i.e. more than nine row pins, some of which are C14 and/or C15) (the "or" is new, determined by trying different variations of row_pins)
• Newly discovered situation: Like the code above, but swapped row_pins and column_pins (and thus also columns_to_anodes=False). At first, everything works fine, but then quickly pressing eight keys simultaneously stops key_matrix from detecting any further events.

I'm using a board which (unlike the BlackPill) doesn't have a crystal at those pins. In mpconfigboard.h, I have #define BOARD_HAS_LOW_SPEED_CRYSTAL (0) (which as far as I understand means that CircuitPython will assume that there is no low-speed crystal).

So why do pins C14 and C15 affect the behavior of other pins? Does CircuitPython give them some additional special treatment except for the never_reset_pin_number() that I commented out? Or does the MCU consider certain signals at those pins as signals coming from a crystal (despite #define BOARD_HAS_LOW_SPEED_CRYSTAL (0)), and partially stops working? (KeyMatrix events don't get detected, but other commands such as print('foo') still work though.)

[rgbpcb]

Possibly unrelated: When leaving out pins C14 and C15 from the key matrix, a much quicker press of many keys at once (quicker and more keys than the aforementioned "quickly pressing eight keys simultaneously") can stop the detection of any further events. Even if row_pins and column_pins each are fewer than ten pins. Increasing max_events doesn't help. Weirdly though, decreasing it to max_events=8 seems to prevent this problem when pins C14 and C15 are left out. So it seems that keypad can't handle a large event queue, for example the default max_events=64.

On the other hand, when pins C14 and C15 are part of the key matrix, max_events=8 doesn't prevent the problem. That's why I wrote "possibly unrelated" above.

[dhalbert]

Maybe related to #6553