Notification callback causes exception in gi

Quick workaround, in case this is blocking you:

    notification.add_action(
        "action_id",
        "action_label",
        action_callback,
        (None,),
        None
    )

15:28 < asac> pitti: is it better from your perspective to implement a workaround in unity8 for the pygi
              regression (compared to just backout pygi changes?)? whats the real fix?
15:28 < pitti> asac: if you need to fix it right now, you can add the workaround (it won't hurt after the fix)
15:29 < pitti> asac: I need to catch simon feltman about this, but he's asleep, so will need to wait until tomorrow
15:30 < asac> pitti: so just reverting the pygi isnt feasible technically? point is that unity8 would have to land
              with a workaround and unity8 is pretty big beast
15:31 < asac> we can do that, but from outside perspective i see both options as valid and would choose whatever
              is less work :)
15:31 < asac> (assuming that we solve this problem upstream anyway)
15:31 < pitti> asac: certainly it's feasible
15:32 -!- pmcgowan [~pat@91.189.91.3] has joined #ubuntu-touch
15:34 < asac> pitti: so is the change/workaround in unity8 the right thing to do? if its the right thing anyway, i
              would go the unity8 route; otherwise only if its more work than a revert upload ..
15:34 * asac still wonders how python is hitting us at all in phone
15:34 < asac> thought we dont do python there :/
15:34 < pitti> yeah, I thought that, too
15:34 < pitti> asac: currently testing a reversion of the corresponding commit in pygobject
15:35 < asac> cool

This bug was fixed in the package pygobject - 3.11.3-0ubuntu2

---------------
pygobject (3.11.3-0ubuntu2) trusty; urgency=medium

  * Add revert_userdata_varargs.patch: Revert upstream commit 9456e83 to fix
    regression with "None" userdata callback arguments in some cases.
    (LP: #1268578)
 -- Martin Pitt <email address hidden> Mon, 13 Jan 2014 15:39:58 +0100

This has been fixed properly in pygobject 3.11.4, and I just committed a test case for this to pygobject upstream master.

This bug was fixed in the package unity8 - 7.84+14.04.20140128-0ubuntu1

---------------
unity8 (7.84+14.04.20140128-0ubuntu1) trusty; urgency=low

  [ Michal Hruby ]
  * Added unity-scope-tool, which will help when developing scopes.

  [ Michał Sawicz ]
  * Added unity-scope-tool, which will help when developing scopes.
  * Use full DashContent, not just GenericScopeView in ScopeTool.qml.
  * Bring Card and CardHeader over from new-scopes.
  * Work around bug #1268578. (LP: #1268578)
  * Drop unnecessary version dependencies.
  * Return null instead of undefined from findChild and
    findInvisibleChild.
  * Fix CardHeader and Card heights (empty Label does have non-zero
    height apparently). Also improve test robustness and reduce future
    diffs.

  [ Albert Astals ]
  * Do not assert if the item we are removing was not created yet
    (because e.g. it's not in the viewport).
  * Position correctly the pointer of the search history box .
  * Make test_filter_expand_expand less unstable in CI VMs Make sure
    header0 is the header0 we want to click On the CI VM stuff is a bit
    slower than on real hw and we were clicking in the wrong place.
  * Add TabBar to the Dash header navigation Changes this comes with: *
    DashBar at the bottom is gone * PageHeader doesn't have a Label
    anymore, it has the childItem property where you add which thing it
    has to contain * New: PageHeaderLabel mimics the old behaviour of
    PageHeader * The header of the LVWPH of GenericScopeView is now fake
    and only used for positioning. There is a single global floating
    header in DashContent (which is a PageHeader with a TabBar as
    childItem) * The GenericScopeView previewLoader and OpenEffect have
    been also moved to the DashContent so that the openEffect includes
    the floating header in the "animation" .
  * Introduce the HorizontalJournal.
  * If there are no items m_firstVisibleIndex has to be -1 .
  * Add some more documentation about tests to the CODING file.
  * Fixes to the journal cmake tests code * Output to the correct
    filename for the test * Don't output stuff from the tryXYZ targets.
  * Adapt to findChild return value changes .
  * Organic Grid for the Dash View.
  * Misc journal fixes Don't init *modelIndex to INT_MAX Makes no sense
    since we're not doing any qMin and the calling function also accepts
    any index >= 0 as valid so in some cases it may end up wanting to
    create an index that doesn't exist Don't refill if height() < 0,
    that gives bad ranges for from/to and the code gets confused .

  [ Michał Karnicki ]
  * Fix grid view column count.
  * Add test for minimum number of items in a carousel.

  [ Allan LeSage ]
  * Add stubs for indicators autopilot tests.

  [ Andrea Cimitan ]
  * Avoid input falling through notifications onto surfaces below, thus
    fixing LP: #1257312. (LP: #1257312)

  [ Leo Arias ]
  * Close the Touch devices after the tests. (LP: #1267600)
  * Added methods to scroll to other scopes on autopilot tests.
  * Added autopilot helpers for the app scope and the app preview.
    Install the fake scopes in order to use them on the tests. (LP:
    #1269114)
  * On autopilot hel...