ubuntu 16.04 , nvidia-docker installed, a tensorflow container running, python 2.7
i want to run a simple python code inside the container. shown as below
from tkinter import *
master = Tk()
canvas_width = 80
canvas_height = 40
w = Canvas(master,
width=canvas_width,
height=canvas_height)
w.pack()
y = int(canvas_height / 2)
w.create_line(0, y, canvas_width, y, fill="#476042")
mainloop()
but when i run this, this error comes
_tkinter.TclError: no display name and no $DISPLAY environment variable
and according to this site https://github.com/moby/moby/issues/8838 i committed my container to an image, and run it again with the -e flag:
-e DISPLAY=:0.0
but here comes the error:
_tkinter.TclError: couldn't connect to display :0.0
i'm not using ssh to login the container. could anyone can give me advice about it?
This is because the container couldn't access the x11 socket of the host. so when doing the docker run, need to include these two flag.
-v /tmp/.X11-unix:/tmp/.X11-unix
-e DISPLAY=unix$DISPLAY
and after this, we need to do another operation. because the default settings of X11 only allows local users to print. so we need to change this to all users.
$ sudo apt-get install x11-xserver-utils
$ xhost +
then the problem solved. ^ ^
_tkinter.TclError: couldn't connect to display "localhost:0.0"
i have same error message, solved. (at least the error message gone!)
export MPLBACKEND='Agg'