I'm currently working on wdspec tests for window manipulation over on bug 1521028. When running the fullscreen tests in headless mode I can see the following response via Marionette:

0:23.85 pid:68389 1547843043945 Marionette DEBUG 0 -> [0,32,"WebDriver:FullscreenWindow",{}]
[..]
0:24.11 pid:68389 1547843044203 Marionette DEBUG 0 <- [1,32,null,{"x":100,"y":100,"width":1366,"height":768}]

That means that the window itself has the maximum dimension of the screen, but the window is not move to the top-left corner of the screen.

This blocks us from enabling the wdspec tests in headless mode.

Note that the original rect of the window were: x=100, y=100, width=800, height=600.

I would expect a position of x=0, y=0 when going into fullscreen mode. But the window stays at the original position.

Have a little test case I can run?

Sure, I will attach in a moment.

The test is passing for me on linux w/ headless:

1548454445878 Marionette DEBUG 3 -> [0,5,"WebDriver:FullscreenWindow",{}]
1548454446152 Marionette DEBUG 3 <- [1,5,null,{"x":0,"y":0,"width":1366,"height":768}]
1548454446156 Marionette DEBUG 3 -> [0,6,"WebDriver:DeleteSession",{}]
1548454446157 Marionette DEBUG 3 <- [1,6,null,{"value":null}]
0:04.26 TEST_END: PASS

Now I also tested with Linux, and indeed it works fine there. But it still fails on MacOS:

1548664850512 Marionette DEBUG 4 -> [0,2,"WebDriver:FullscreenWindow",{}]
1548664850798 Marionette DEBUG 4 <- [1,2,null,{"x":4,"y":4,"width":1366,"height":768}]

It's still a problem on MacOS and maybe related to bug 1563161?