[Bug] ev3-micropython V2 does not auto register EV3-color-sensor

[BertLindeman]

Describe the bug
On ev3-micropython (the system) EV3 and NXT color sensors are not auto-detected.
Hope I am entering this on the correct github repository.
And hope this is not too much info

To reproduce
Steps to reproduce the behavior:

1. Running ev3-micropython 2.0
2. on a terminal type the command "dmesg -w"
3. connect the colorsensor to a sensor port, e.g. S3
4. Look at the dmesg output of switching varipus color sensors (last is nxt-color sensor:
   [Sat Jul 18 00:23:13 2020] lego-port port2: Removed device 'ev3-ports:in3:ev3-uart-host'
[Sat Jul 18 00:23:28 2020] lego-port port2: Added new device 'ev3-ports:in3:ev3-uart-host'
[Sat Jul 18 00:23:50 2020] lego-port port2: Removed device 'ev3-ports:in3:ev3-uart-host'
[Sat Jul 18 00:23:55 2020] lego-port port2: Added new device 'ev3-ports:in3:ev3-uart-host'
[Sat Jul 18 00:24:00 2020] lego-port port2: Removed device 'ev3-ports:in3:ev3-uart-host'
[Sat Jul 18 00:24:03 2020] lego-port port2: Added new device 'ev3-ports:in3:lego-nxt-color'

Note missing: "lego-sensor sensor2: Registered 'lego-ev3-color' on 'ev3-ports:in3'"

5. ls -l /sys/class/lego-sensor/
   and see no new sensor has been registered there.

Expected behavior
See the color sensor in the /sys/ filesystem AND have the sensor connected in my micropython program

Screenshots
not now

Desktop:

• OS and version: [Windows 10 Home 1909]
• Connection Type: [Wi-Fi]
• Software: [Visual Studio Code and PuTTY]

Robot:

• Device: EV3
• Versions:
  Image file: ev3-micropython-v2.0.0-sd-card-image
  Kernel version: 4.14.117-ev3dev-2.3.5-ev3
  Brickman: 0.10.3
  BogoMIPS: 148.88
  Bluetooth: 4.0
  Board: board0
  BOARD_INFO_HW_REV=7
  BOARD_INFO_MODEL=LEGO MINDSTORMS EV3
  BOARD_INFO_ROM_REV=6
  BOARD_INFO_SERIAL_NUM=0016535166E2
  BOARD_INFO_TYPE=main

debugging done so far
Switched back to ev3dev with the same sensors / motors on the same ports:
Welcome to ev3dev GNU/Linux 9.12 (stretch) GNU/Linux
Kernel: 4.14.117-ev3dev-2.3.5-ev3 armv5tejl
EV3DEV image: ev3dev-stretch-ev3-generic-2018-11-13
and did the same with success.
dmesg tail:
[Fri Jul 17 20:19:53 2020] lego-port port2: Added new device 'ev3-ports:in3:ev3-uart-host'
[Fri Jul 17 20:19:56 2020] lego-sensor sensor2: Registered 'lego-ev3-color' on 'ev3-ports:in3'.

and:
showsensors:
/sys/class/lego-sensor/sensor2
sensor 1 driver_name: lego-ev3-color on port ev3-ports:in3
address: ev3-ports:in3
mode: COL-REFLECT
modes: COL-REFLECT COL-AMBIENT COL-COLOR REF-RAW RGB-RAW COL-CAL
commands:
decimals: 0
num_values: 1
value0: 1
units: pct
uevent:
LEGO_DRIVER_NAME=lego-ev3-color
LEGO_ADDRESS=ev3-ports:in3

Background Information
Trying to program the advanced color sorter of MOC-5615 - Optimized Colour Sorter
The color sensor (both EV3 and NXT) are recognized as seen in dmesg, but not instantiated in the /sys/ filesystem.
And thus not found by micropython. (My conclusion)

The same does work on ev3dev

[dlech]

The NXT color sensor will never appear in /sys/class/lego-sensor/. ev3dev can detect it but there is no driver. It works in Pybricks by using the raw mode of the port.

Not sure why the EV3 color sensor is not syncing up though. There really isn't much difference between the two SD card images.

You could just use the working ev3dev SD card. (sudo apt update && sudo apt install pybricks-micropython to get the latest Pybricks)

[BertLindeman]

@dlech Yes, the ev3dev image does work, so that would be my circumvention.
The NXT color sensor was just an attempt to get some more debug info.

If there is anything (easy ;-) ) I can do to debug, let me know.

[laurensvalk]

Can you try both sensors on different ports?

When you use nxtdevices.ColorSensor, it disables autodetect because it puts the sensor in raw mode. So you will not see anything in the device browser menu after this. Auto id is automatically re-enabled the next time you use ev3devices.ColorSensor.

Other ports are unaffected, so you should still be able to see the EV3 color sensor there.

In other words, I’m wondering if there is perhaps no issue :)

[BertLindeman]

Can you try both sensors on different ports?

@laurens I would forget about the NXT color sensor in this case, unless it os important for debugging. In this case I no not need it.

Did a reboot (pybricks SD card) with two EV3 color sensors on S2 and S3 to get a stable situation.

dmesg|grep -i ev3-port
shows:

[Sat Jul 18 09:43:59 2020] lego-port port0: Registered 'ev3-ports:in1' on 'ev3-ports:in1'.
[Sat Jul 18 09:43:59 2020] lego-port port1: Registered 'ev3-ports:in2' on 'ev3-ports:in2'.
[Sat Jul 18 09:43:59 2020] lego-port port2: Registered 'ev3-ports:in3' on 'ev3-ports:in3'.
[Sat Jul 18 09:43:59 2020] lego-port port3: Registered 'ev3-ports:in4' on 'ev3-ports:in4'.
[Sat Jul 18 09:43:59 2020] lego-port port4: Registered 'ev3-ports:outA' on 'ev3-ports:outA'.
[Sat Jul 18 09:43:59 2020] lego-port port5: Registered 'ev3-ports:outB' on 'ev3-ports:outB'.
[Sat Jul 18 09:43:59 2020] lego-port port6: Registered 'ev3-ports:outC' on 'ev3-ports:outC'.
[Sat Jul 18 09:43:59 2020] lego-port port7: Registered 'ev3-ports:outD' on 'ev3-ports:outD'.
[Sat Jul 18 09:43:59 2020] lego-port port0: Added new device 'ev3-ports:in1:lego-ev3-touch'
[Sat Jul 18 09:43:59 2020] tacho-motor motor0: Registered 'lego-ev3-l-motor' on 'ev3-ports:outA'.
[Sat Jul 18 09:43:59 2020] lego-port port4: Added new device 'ev3-ports:outA:lego-ev3-l-motor'
[Sat Jul 18 09:43:59 2020] tacho-motor motor1: Registered 'lego-ev3-m-motor' on 'ev3-ports:outD'.
[Sat Jul 18 09:43:59 2020] lego-port port7: Added new device 'ev3-ports:outD:lego-ev3-m-motor'
[Sat Jul 18 09:44:00 2020] lego-port port1: Added new device 'ev3-ports:in2:ev3-uart-host'
[Sat Jul 18 09:44:00 2020] lego-port port2: Added new device 'ev3-ports:in3:ev3-uart-host'
[Sat Jul 18 09:44:46 2020] lego-sensor sensor0: Registered 'lego-ev3-touch' on 'ev3-ports:in1'.

So the touch sensor on S1 is registered and the ev3-color sensors are not.

Tried to look at udev rules, but could not find significant diff between ev3dev and pybricks SD-card images.
But I am no expert on this (far from)

[BertLindeman]

Maybe a red herring, but I saw this (running pybricks SD) in journalctl -r

Jul 18 10:46:02 ev3dev systemd[1]: Failed to start LEGO MINDSTORMS EV3 UART Sensors.
Jul 18 10:46:02 ev3dev systemd[1]: ev3-uart@ev3x2dports:in2.service: Failed with result 'exit-code'.
Jul 18 10:46:02 ev3dev systemd[1]: ev3-uart@ev3x2dports:in2.service: Control process exited, code=exited, status=1/FAILURE
Jul 18 10:46:02 ev3dev ldattach[1414]: ldattach: cannot open /dev/tty_ev3x2dports:in2: No such file or directory
Jul 18 10:46:02 ev3dev systemd[1]: Starting LEGO MINDSTORMS EV3 UART Sensors...
Jul 18 10:46:02 ev3dev systemd[1]: Unnecessary job for /sys/subsystem/lego/devices/ev3_uart_host_ev3x2dports:in2 was removed.
Jul 18 10:45:59 ev3dev kernel: lego-port port1: Added new device 'ev3-ports:in2:ev3-uart-host'

[laurensvalk]

Thanks for clarifying that this is only about the EV3 sensors. In that case, yes, it might be an issue with ev3dev. We've been seeing quite a few issues with auto-detect lately, both motors and sensors.

• Does this happen every time? (I've only seen it occasionally and could not consistently reproduce.)
• Does it persist after you unplug the cable and plug it back in?
• Are you certain it happens only on the LEGO EV3 MicroPython version of the image? Both are just ev3dev and with the same kernel. It could be the case, but it could be a coincidence. Retrying a couple times might shed some light on this.

We appreciate the input. Thanks!

[BertLindeman]

@laurensvalk

Does this happen every time? (I've only seen it occasionally and could not consistently reproduce.)

Yes constantly

Does it persist after you unplug the cable and plug it back in?

Yes

Are you certain it happens only on the LEGO EV3 MicroPython version of the image? Both are just ev3dev and with the same kernel. It could be the case, but it could be a coincidence. Retrying a couple times might shed some light on this.

Sure Laurens, ONLY the pybricks image.

Maybe a hint in my previous comment.

Will re-boot the ev3dev image and re-try a couple of thime to be sure the error is not in that image.

Thanks for your to the point remarks.

[laurensvalk]

That's very interesting, thanks---Sorry for those questions, I just wanted to rule out a few possibilities we've encountered earlier.