The problem seems to be that the GstClockID C type is just "gpointer" (thus no GObject).  To solve this, one needs to implement a new python object (similiar to gst.Iterator).  I'm looking into this.  If I can get it to work correctly, I will provide a patch.

It might be MUCH easier to actually make it a GBoxed.
It won't change the C typedef (gpointer), but it will add a GType, with some copy, new and free methods registered with it.
That should be the proper way.
If there are still some non-wrappable structures/pointers in GStreamer, they should have a declared GBoxed GType to go along with it.

Custom code is needed anyways, but it seems to be even less complex than e.g. gst.Iterator.  Implementing ClockIDs as (non-gobject) python objects is rather straightforward.  What's driving me nuts is the code generator.

does this need a a GStreamer -core addition, like add a GType, or so?

No, I have a pure python solution that works well and is almost finished.

Created attachment 78033 [details] [review]
Python binding for ClockID objects

Description of changes:

        * gst/common.h:
        Add new PyGstClockID object type structure.
        * gst/gstmodule.c:
        Module setup for new type PyGstClockID_Type.
        * gst/gst.override: (_wrap_gst_clock_new_single_shot_id,
        _wrap_gst_clock_new_periodic_id):
        Add overrides for ClockID factory functions.  These create objects
        of the new type, correctly verify time arguments and do not let
        threads run needlessly.
        * gst/Makefile.am:
        Add new file pygstclockid.c containing the gst.ClockID object type
        implementation.
        * gst/pygstclockid.c: (set_exception_from_clock_return,
        pygst_clock_id_async_callback, pygst_clock_id_compare,
        pygst_clock_id_repr, pygst_clock_id_hash, pygst_clock_id_traverse,
        pygst_clock_id_clear, pygst_clock_id__new__, pygst_clock_id_new,
        pygst_clock_id_dealloc, pygst_clock_id_get_time,
        pygst_clock_id_wait, pygst_clock_id_wait_async,
        pygst_clock_id_unschedule, _PyGstClockID_methods,
        PyGstClockID_Type):
        Implementation of new object type gst.ClockID.
        * pygstexception.c, pygstexception.h: (PyGstExc_ClockError,
        PyGstExc_ClockIDUnscheduled, pygst_exceptions_register_classes):
        Add new exceptions gst.ClockError and gst.ClockIDUnscheduled.
        * testsuite/Makefile.am:
        Add testsuite for SystemClock/ClockID objects.
        * testsuite/test_clock.py (seconds, SystemClockTestBase, .setUp,
        .tearDown, ClockIDTest, .testGetTime, .testClockTimeNone,
        .testWait1, .testWait2, .testWaitAsync, .testInvalidWait,
        .testInvalidWaitAsync, .testAbortWait, .testAbortWaitAsync,
        .testComparison, .testInvalidConstructor, .testRefCycle):
        Testsuite for gst.ClockID objects.

Created attachment 78381 [details] [review]
Python binding for ClockID objects (2)

Same patch as before, but also removes the now unneeded arg type mapping from GstClockID to gpointer:

        * gst/arg-types.py:
        Remove unneeded arg type override for GstClockID types.

Nice patch.

You could make it even smaller (and easier to maintain), by not writing specific overrides for creating clocks, but just defining a new argtype for GstClockId in gst/arg-types.py (like it is done for GstIterator, and like you did for string arrays in another patch you proposed)... unless I missed something in the patch that doesn't allow doing this.

That also occured to me afterwards, but like I said, the overrides are needed anyways since we have to verify the time arguments and raise ValueErrors if they are invalid.

In the long run, it makes lots of sense to add an argument attribute (or what it's called) valid-clock-time, similar to the existing null-ok etc.  This would allow us to define the methods like this:

(define-method new_periodic_id
  (of-object "GstClock")
  (c-name "gst_clock_new_periodic_id")
  (return-type "GstClockID")
  (parameters
    '("GstClockTime" "start_time" (valid-clock-time))
    '("GstClockTime" "interval" (valid-clock-time not-null)
  )
)

But this can be done as separate work, as it is relevant to lots of functions.  It's a nice way to get rid of some overrides and add better checking to the functions that miss it.

Doesn't apply against CVS HEAD.

Any chance we can revive this? I really need it for my application...

*** Bug 562280 has been marked as a duplicate of this bug. ***

Created attachment 129369 [details] [review]
updated patch, works against HEAD of right now

In [4]: i = c.new_single_shot_id(2414)
/usr/bin/python: symbol lookup error: /usr/lib/python2.6/dist-packages/gst-0.10/gst/_gst.so: undefined symbol: pygst_clock_id_new

Also:

In [2]: c = gst.ClockID()
Segmentation fault

Can you please pick this back up? I really need it for my app!

Closing this bug now, gst-python is only an extension module to pygi now and this bug doesn't make much sense anymore in this context.