test: Add desktopui_WindowManagerFocusNewWindows.

test: Add desktopui_WindowManagerFocusNewWindows.

This uses the new Python version of autox to open
a few windows and check that the window manager
moves them onscreen and focuses them.  Note that
it doesn't actually check what gets composited,
though (although that should soon be pretty
straightforward to add).

BUG=chromium-os:2041, chromium-os:2229
TEST=ran it on x86-generic

[Daniel Erat, 2010-03-25 23:25:05]

I also need to deal with the possibility of not being logged in initially, I
guess.  Do you think that I should call the existing stuff that uses the C++
version of autox?

http://codereview.chromium.org/1371003/diff/1/3
File
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py
(right):

http://codereview.chromium.org/1371003/diff/1/3#newcode50
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py:50:
os.symlink('/home/chronos/.Xauthority', '/root/.Xauthority')
I'm not going to check this part in.  I'm still looking for a better solution
(this is actually a non-solution, as it requires that the root directory be
remounted read-write).  I might need to patch python-xlib to support overriding
the auth file.

http://codereview.chromium.org/1371003/diff/1/3#newcode51
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py:51:
self.autox = autox.AutoX(':0')
I'll find something better for this too.  Maybe we should set DISPLAY to :0 when
running tests.

[Daniel Erat, 2010-03-26 01:27:00]

Okay, this is ready now.

http://codereview.chromium.org/1371003/diff/4001/1007
File
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py
(right):

http://codereview.chromium.org/1371003/diff/4001/1007#newcode61
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py:61:
# For now, we just sleep a really long time.
I opened http://crosbug.com/2264 to track this.

[kmixter1, 2010-03-26 02:30:08]

http://codereview.chromium.org/1371003/diff/4001/1007
File
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py
(right):

http://codereview.chromium.org/1371003/diff/4001/1007#newcode61
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py:61:
# For now, we just sleep a really long time.
On 2010/03/26 01:27:00, Daniel Erat wrote:
> I opened http://crosbug.com/2264 to track this.

Can we just wait for cryptohome to be mounted or is WM starting asynchronous? 
This seemed adequate when I ran into this on a slower HDD device:

http://codereview.chromium.org/1213005

[Daniel Erat, 2010-03-26 02:37:28]

It's not just the WM starting (I already watch for its process); it's
Chrome starting and its initial window being moved onscreen by the
window manager, and yeah, this happens asynchronously.  Waiting for
cryptohome to be mounted is adequate for the test that you linked to
because that's all that the test needs -- trying to figure out if the
UI is completely set up will probably require some cooperation from
Chrome and/or the window manager.  I initially tried to just poll for
a window being onscreen using autox, but it looks like that's still
flaky.  Longer-term, I can just make the window manager touch a file
or something after it sees a Chrome window.

On Thu, Mar 25, 2010 at 7:30 PM,  <kmixter@chromium.org> wrote:
> http://codereview.chromium.org/1371003/diff/4001/1007
> File
>
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py
> (right):
>
> http://codereview.chromium.org/1371003/diff/4001/1007#newcode61
>
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py:61:
> # For now, we just sleep a really long time.
> On 2010/03/26 01:27:00, Daniel Erat wrote:
>>
>> I opened http://crosbug.com/2264 to track this.
>
> Can we just wait for cryptohome to be mounted or is WM starting
> asynchronous?  This seemed adequate when I ran into this on a slower HDD
> device:
>
> http://codereview.chromium.org/1213005
>
> http://codereview.chromium.org/1371003
>

[Chris Masone, 2010-03-26 07:03:43]

lgtm

http://codereview.chromium.org/1371003/diff/4001/1007
File
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py
(right):

http://codereview.chromium.org/1371003/diff/4001/1007#newcode64
client/site_tests/desktopui_WindowManagerFocusNewWindows/desktopui_WindowManagerFocusNewWindows.py:64:
# TODO: Set these in a single, standard place for all tests.
Perhaps autox.AutoX() can do it, and take params for these values.  The ones
used here could be the default?  We can store them in chromeos_constants.

http://codereview.chromium.org/1371003/diff/4001/1008
File
client/site_tests/login_ChromeProfileSanitary/login_ChromeProfileSanitary.py
(right):

http://codereview.chromium.org/1371003/diff/4001/1008#newcode59
client/site_tests/login_ChromeProfileSanitary/login_ChromeProfileSanitary.py:59:
raise error.TestError('Could not login')
Thank you for fixing this