Serial monitor gives corrupted output

[cmaglie]

Describe the problem

To reproduce

This is very difficult to reproduce and, after a lot of trial and error, we found a combination of HW and SW that is more likely to reproduce it:

• A macOS machine
• A board that mounts a CP2102 chip as a USB-2-serial converter

Here are the instructions to reproduce:

1. Upload the following sketch to the board (note it is configured for 9600 baud):

   void setup() {
     Serial.begin(9600);
   }
   void loop() {
     Serial.println("Hello, world!");
     delay(1000);
   }

2. Open the Cloud Editor Serial Monitor
   ✅ The serial output is printed without corruption (Cloud Editor Serial Monitor always starts configured for 9600 baud).
3. Select "115200" from the baud rate menu.
4. Select "9600" from the baud rate menu.
   ❌ The serial output is corrupted with swapped or missing characters, as described by the users.

The fault only occurs when changing the baud rate in the Serial Monitor. It is not specific to the 9600 baud rate. For example, I can also reproduce the fault with this more simplified procedure:

1. Upload the following sketch to the board (note that I configured it for 115200 baud instead of 9600 this time):

   void setup() {
     Serial.begin(115200);
   }
   void loop() {
     Serial.println("Hello, world!");
     delay(1000);
   }

2. Open the Cloud Editor Serial Monitor.
   ✅ The serial output is corrupted as expected because the Serial Monitor is configured for 9600 baud.
3. Select "115200" from the baud rate menu.
   ❌ The serial output is corrupted with swapped or missing characters, as described by the users.

Expected behavior

The Serial Monitor shows uncorrupted output.

Arduino Create Agent version

1.7.0

Operating system

macOS

Operating system version

Sequoia

Browser

Chrome

Browser version

135.0.7049.115

Additional context

Additional reports

• https://forum.arduino.cc/t/arduino-uno-with-garbled-serial/1354153

Issue checklist

• I searched for previous reports in the issue tracker
• I verified the problem still occurs when using the latest version
• My report contains all necessary details

[cmaglie]

Other information from @per1234:

When I examine the commands that are produced by the above procedure in the Cloud Agent Debug Console, I notice that the 3rd step causes multiple open commands, instead of a single open step as we would expect:

open /dev/cu.usbserial-0001 9600 timed

{
  "Cmd": "Open",
  "Desc": "Got register/open on port.",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 9600,
  "BufferType": "timed"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "\u0000��\u0000�\u0000���\u0000 \u00017\u0017$Nu%���n9�\u0018�\b�"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "�"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "["
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "+"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "�"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "K"
}

close /dev/cu.usbserial-0001

Closing serial port /dev/cu.usbserial-0001

Shutting down reader on /dev/cu.usbserial-0001

{
  "Cmd": "Close",
  "Desc": "Got unregister/close on port.",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 9600
}

Shutting down writer on /dev/cu.usbserial-0001

writerBuffered just got closed. make sure you make a new one. port:/dev/cu.usbserial-0001

open /dev/cu.usbserial-0001 115200 timed

open /dev/cu.usbserial-0001 115200 timed

{
  "Cmd": "Open",
  "Desc": "Got register/open on port.",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 115200,
  "BufferType": "timed"
}

{
  "Cmd": "Open",
  "Desc": "Got register/open on port.",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 115200,
  "BufferType": "timed"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "�\u0003\u0000]�R\u001ep51��7q!\u00185�`x400805cc\r\n"
}

open /dev/cu.usbserial-0001 115200 timed

{
  "Cmd": "OpenFail",
  "Desc": "Error opening port. Serial port busy",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 115200
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "H\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "ello, world!"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "H"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "ello, world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "H"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "ello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "H"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "ello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hworld!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "ello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "world!\r\n"
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "ello, "
}

{
  "P": "/dev/cu.usbserial-0001",
  "D": "Hworld!\r\n"
}

The third of those open commands (the second redundant command) correctly failed:
open /dev/cu.usbserial-0001 115200 timed

{
  "Cmd": "OpenFail",
  "Desc": "Error opening port. Serial port busy",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 115200
}

but the first two (the expected command, and the first redundant one) went through:

open /dev/cu.usbserial-0001 115200 timed

open /dev/cu.usbserial-0001 115200 timed

{
  "Cmd": "Open",
  "Desc": "Got register/open on port.",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 115200,
  "BufferType": "timed"
}

{
  "Cmd": "Open",
  "Desc": "Got register/open on port.",
  "Port": "/dev/cu.usbserial-0001",
  "Baud": 115200,
  "BufferType": "timed"
}

I was not able to reproduce the fault by sending the same commands manually via the Cloud Agent debug console. The redundant open commands always correctly failed, no matter how fast I tried to send the sequential commands.

• I reproduced the fault on a macOS Sequoia machine
• I am using the native macOS support for the CP2102. I have never manually installed any drivers on the machine.
• I am not able to reproduce the fault on a Windows 11 machine
• I am not able to reproduce the fault with an Adafruit Feather ESP32 board that has the related Silicon Labs CP2104 chip.
• I am not able to reproduce the fault with a Nano derivative board that has the CH340.
• I am not able to reproduce the fault with an official UNO R3.
• I am able to reproduce the fault if I use 57600 baud instead of the 115200 arbitrarily selected for the demonstration procedures I described above.
• I am able to reproduce the fault with the board connected directly to the USB socket on the computer, and with it connected via a USB hub.