File access won't be closed on /dev/i2c

[toefde]

• Input source: HDMI
• Output device: Raspberry Pi 4
• Operating system: Raspberry OS
• Python version: 3
• monitorcontrol version: 3.0.2

Hi, I've got a problem with the python library.
Everytime I execute monitorcontrol.get_monitors() it accesses the file (in my case) /dev/i2c-21 with a new file descriptor.
I need to execute it every second and after a few hundred iterations my script crashes with [Errno 24] Too many open files. Because there is a limit of open files on the system (Yes, I could increase the limit, but it won't help) is it possible to dispose whatever the library does there? I need to somehow evade this problem.

Thanks for the help

[newAM]

I'm pretty busy this weekend, but I can try to take a look next weekend. This was the first python library I wrote, it's likely that there's just something sub-optimal somewhere :)

Do you have some simple code I can use to reproduce this?

[toefde]

Yes of course.

• Run python3 on the terminal
• Open a second terminal and get the pid of python3 ps -a
• With sudo lsof -p pidofpython3 there is no /dev/i2c device in access
• Run in the python terminal:
  • import monitorcontrol
  • test = monitorcontrol.get_monitors()
• Now sudo lsof -p pidofpython3 there is one /dev/i2c device in access
• And if you run test = monitorcontrol.get_monitors() in the python terminal again, there will be more /dev/i2c devices in access

The limit of open files will be reached sooner or later. But keep in mind the limit depends on if the python script is started as a service or started automatically on startup (standard open file limit is 1024 on Raspberry OS on a RPi 4 Model B 2GB). There are more differences if I execute this manually over ssh in the Visual Studio Code terminal the script won't crash. I don't know the difference there yet.

So let's grab the problem at the root and try to avoid the multiple open i2c files. :)

Huge thanks for the quick support

[newAM]

Thanks for that reproduction, that was very helpful!

This is indeed a bug. When exceptions were generated, which occurred for non-monitor devices on the i2c bus __exit__ never got called and the file descriptors never got closed.

I normally wait for feedback on fixes, but I'm pretty confident that #252 fixes this, so I'll merge that and cut a release. If that doesn't fix it let me know and I can do another release!