Created attachment 59674 [details] [review]
Proposed patch to make probes on the target pad if it's a ghostpad

*** Bug 325259 has been marked as a duplicate of this bug. ***

the ghostpad should behave like any other pad so the fix should be something local to the ghostpads. I might have a clue what's wrong here.

Created attachment 59716 [details] [review]
Apply the same fix gst_pad_set_blocked

Created attachment 59740 [details] [review]
Same patch but without leaking the target pad refcount

Fixed in CVS HEAD

reopening as we must find something that is more transparent.

I'm thinking about a notification when a probe/block is installed.

It doesn't work if the target of the ghostpad is also a ghostpad.

Created attachment 64873 [details] [review]
multi-ghostpad fix for ghostpad

This is the proposed patch for gst_pad_[add|remove]_*_probe().

Instead of assuming that there can only be ghostpads of real pad, it will recursively got up the targets until it reaches a non-ghostpad.

Created attachment 64875 [details] [review]
updated version, handles gst_pad_set_blocked_async() too

We also need to handle the case for gst_pad_set_blocked_async().

Created attachment 64877 [details] [review]
Updated version, with a unit test to show the fix

Better with a unit test.

The patch is incorrect wrt refcounts.

  while (GST_IS_GHOST_PAD (pad)) {
    if (was_ghost)
      gst_object_unref (pad);
    pad = gst_ghost_pad_get_target (GST_GHOST_PAD (pad));

You only have one reference on pad in the was_ghost==TRUE case, but you use its value after dropping your reference. You would need to make a temp variable and unref it after the get_target.

However this whole idea of including gstghostpad.h from within gstpad.c is incorrect. GstPad should supply an interface general enough that gstghostpad can implement this functionality on its own, without having to have gstpad.c explicitly call out to it.

I created a new bug for the probe on ghostpad of ghostpad issue : 341029.

Marking this bug as a blocker of that one.

2006-07-11  Edward Hervey  <edward@fluendo.com>

        * gst/gstbin.c: (activate_pads),
        (iterator_activate_fold_with_resync), (gst_bin_src_pads_activate),
        (gst_bin_change_state_func):
        (de)activate src pads before calling state_change on the childs.
        This is to avoid the case where a src ghostpad is blocked (holding the
        stream lock), which would block the deactivation of the ghostpad's
        target pad.
        * gst/gstghostpad.c: (gst_proxy_pad_do_query_type),
        (gst_proxy_pad_do_event), (gst_proxy_pad_do_query),
        (gst_proxy_pad_do_internal_link), (gst_proxy_pad_do_bufferalloc),
        (gst_proxy_pad_do_chain), (gst_proxy_pad_do_getrange),
        (gst_proxy_pad_do_checkgetrange), (gst_proxy_pad_do_getcaps),
        (gst_proxy_pad_do_acceptcaps), (gst_proxy_pad_do_fixatecaps),
        (gst_proxy_pad_do_setcaps), (gst_proxy_pad_set_target_unlocked),
        (gst_proxy_pad_set_target), (gst_proxy_pad_get_internal),
        (gst_proxy_pad_dispose), (gst_proxy_pad_init),
        (gst_ghost_pad_parent_set), (gst_ghost_pad_parent_unset),
        (gst_ghost_pad_class_init),
        (gst_ghost_pad_internal_do_activate_push),
        (gst_ghost_pad_internal_do_activate_pull),
        (gst_ghost_pad_do_activate_push), (gst_ghost_pad_do_activate_pull),
        (gst_ghost_pad_do_link), (gst_ghost_pad_do_unlink),
        (gst_ghost_pad_dispose), (gst_ghost_pad_new_no_target),
        (gst_ghost_pad_new), (gst_ghost_pad_set_target):
        GhostPads now create their internal GstProxyPad at creation (and not
        when they're linked, as it was being done previously).
        The internal and target pads are linked straight away.
        The data will also travel through the other pad in order to make
        pad blocking and probes non-hackish (the probe/block now really happens
        on the GhostPad and not on the target).
        * gst/gstpad.c: (gst_pad_set_blocked_async),
        (gst_pad_link_prepare), (gst_pad_push_event):
        Remove previous ghostpad cruft.
        * gst/gstutils.c: (gst_pad_add_data_probe),
        (gst_pad_add_event_probe), (gst_pad_add_buffer_probe),
        (gst_pad_remove_data_probe), (gst_pad_remove_event_probe),
        (gst_pad_remove_buffer_probe):
        Remove previous ghost pad cruft.
        Added more detailed debug statements.
        * tests/check/gst/gstghostpad.c: (GST_START_TEST):
        Fix the testsuite for refcounting changes.
        The comments about who has references were correct, but the refcount
        being checked wasn't the same (!?!).