Chrome sessions hang randomly with node-chrome due to DBUS issue

[testphreak]

Meta -

Image(s):
node-chrome. May affect other chrome and firefox images as well.

Image Version:
3.3.1-cesium

Docker version:
17.04.0-ce, build 4845c56

OS:
Ubuntu 16.04.2 LTS

Expected Behavior -

Tests execute as usual without hanging.

Actual Behavior -

@diemol, @ddavison spent a good deal of time last night trying to get to the bottom of the issue. It's related to this change 88584de made by @a-k-g to resolve #358.

Due to this change, none of the node-chrome images after 3.0.1 are working for me and the timeline of the commit confirms that. Would it be possible to add a fix that would work for all images - node-chrome, node-chrome-debug, standalone-chrome and standalone-chrome-debug images?

I passed in DBUS_SESSION_BUS_ADDRESS=/dev/null as an environment variable to node-chrome when starting the containers and all my tests now execute without random hangs with 3.4.0-actinium. It would be better if this variable is set in the image rather than us having to pass it in every time.

I also found out that I am not alone in this issue and several including @pwaller and others have run into this issue recently #87 (comment).

[diemol]

I was able to reproduce it with the Chrome images from the 3.4.0-bismuth release. Summarising, under this use case (open and closing Chrome many times, at least 50):

• node-chrome -> Does not work
• node-chrome-debug -> Works
• standalone-chrome -> Does not work
• standalone-chrome-debug -> Does not work

I am building locally the Dockerfiles now with some changes to see if it works. I'll keep you updated.

[testphreak]

Awesome! Thanks @diemol! Could this issue affect Firefox images as well? I see this line in the dockerfile for NodeFirefox as well.

[diemol]

@testphreak Firefox images seem to work fine. By fine I mean that the test does not hang, it either fails or passes (I am just clarifying it because I am not 100% sure if the latest version of Geckodriver - 0.16 and Firefox 53 are working well, so nothing related to the DBUS_SESSION_BUS_ADDRESS).

[diemol]

@testphreak, I just created a PR for this. Please check it, any help with checking it with your local tests would be great.

[testphreak]

@diemol, your change looks good. Verified your change locally. Thanks for reverting the change back to #182.