I encountered a race condition upon which, even though the udev rules were working and uploads succeed, the serial monitor can not re-establish communications.
If the serial monitor is left open when doing an upload the error shown is:
[...]
Copying data from PC to DFU device
Starting download: [##################################################] finished!
state(8) = dfuMANIFEST-WAIT-RESET, status(0) = No error condition is present
Done!
Resetting USB to switch back to runtime mode
Waiting for /dev/ttyACM0 serial...Done
processing.app.SerialException: Error opening serial port '/dev/ttyACM0'. Try consulting the documentation at http://playground.arduino.cc/Linux/All#Permission
at processing.app.Serial.<init>(Serial.java:145)
at processing.app.Serial.<init>(Serial.java:82)
at processing.app.SerialMonitor$4.<init>(SerialMonitor.java:101)
at processing.app.SerialMonitor.open(SerialMonitor.java:101)
at processing.app.AbstractMonitor.resume(AbstractMonitor.java:104)
at processing.app.Editor.resumeOrCloseSerialMonitor(Editor.java:2218)
at processing.app.Editor.access$2200(Editor.java:79)
at processing.app.Editor$DefaultExportHandler.run(Editor.java:2196)
at java.lang.Thread.run(Thread.java:748)
Error opening serial port '/dev/ttyACM0'. Try consulting the documentation at http://playground.arduino.cc/Linux/All#Permission
I traced it to the device being opened prior to the permissions/ownership changes.
When testing via -c the upload script waits for the character device creation, but at that point the udev rules have not changed the permissions. By testing the readability of the device via the -r test, we wait until the new permissions have been applied.
I imagine this is specific to the Linux kernel and the speed of the machine being used. In my case:
arduino-1.8.5
Linux 4.4.0-98-generic #121-Ubuntu SMP
but the change should work on any system with proper udev rules / permissions.
The INPCK bit is an i_flag bit, not a c_flag bit as originally done
here. I have no idea what setting INPCK in c_flag does on Linux, but
on FreeBSD it caused the test of the c_values after they were set to
fail. This fixes that, and presumably the serial interface will now
actually check the parity of incoming data.
- Fixed 0 byte payload bug, stopping HardWire from sending only addresses to scan for devices
- Fixed I2C bug accessing wrong status register for error flags
- Improved Hardwire endTransmission() return flags to correspond with the actual I2C failure status
- Removed dos endings from dfu-util autogen script breaking compilation