When a bin changes states upwards, and a child fails to change,
any child that was already switched will not be reset to its
original state, leaving its state inconsistent with the bin,
which does not change state due to the failure.
If the state change was from NULL to READY, it means that deleting
this bin will cause those children to be deleted while not in
NULL state, which is a Bad Thing. For other upward changes, it
is less of a problem, as a subsequent switch back to NULL will
cause an actual downwards change on those inconsistent elements,
albeit from the "wrong" state.
We now reset state to the original one when a child fails.
Includes unit test.
https://bugzilla.gnome.org/show_bug.cgi?id=747610
Returning a newly allocated string makes no sense. It's unexpected for a
getter, and also this behaves differently in 0.10, so it would make future
merges harder.
Except for these two places here in core which were updated for the new
semantic, the return value is getting leaked all over the place.
Original commit message from CVS:
* tests/check/generic/states.c: (GST_START_TEST), (states_suite):
Copy the current generic/states example from -base and adapt so
we can use the exact same code everywhere.
Check a STATES_IGNORE_ELEMENTS env var which can be used
to ignore certain element factories for this test, which is
what is being done in -base
* tests/check/Makefile.am:
Mention this environment variable.
Original commit message from CVS:
* gst/gstutils.c: (gst_element_unlink_pads):
Don't leak pad references.
* tests/check/elements/fakesink.c: (GST_START_TEST):
* tests/check/generic/sinks.c: (GST_START_TEST):
* tests/check/generic/states.c: (GST_START_TEST):
* tests/check/gst/gstbin.c: (GST_START_TEST):
* tests/check/gst/gstcaps.c: (GST_START_TEST):
* tests/check/gst/gstelement.c: (GST_START_TEST):
* tests/check/gst/gstghostpad.c: (GST_START_TEST):
* tests/check/gst/gstiterator.c: (GST_START_TEST):
* tests/check/gst/gstvalue.c: (GST_START_TEST):
Fix a bunch of leaks. Make generic/sinks.c
use a bit less cpu by slowing the buffer rate
between fakesrc and fakesink.
Original commit message from CVS:
* check/generic/states.c: (GST_START_TEST), (states_suite):
Add a sleep to ensure elements have a chance to start their
pad tasks before shutdown. Reduces racy test results.
* gst/elements/gstfdsrc.c: (gst_fdsrc_init), (gst_fdsrc_create):
Time out the select every now and then to check for shutdown.
Original commit message from CVS:
* check/Makefile.am:
* check/generic/states.c:
* gst/Makefile.am:
* gst/gst.c:
* gst/gst.h:
* gst/gst_private.h:
* gst/gstelementfactory.c:
* gst/gstindex.c:
* gst/gstinfo.c:
* gst/gstplugin.c:
* gst/gstplugin.h:
* gst/gstpluginfeature.c:
* gst/gstpluginfeature.h:
* gst/gstregistry.c:
* gst/gstregistry.h:
* gst/gstregistrypool.c: remove
* gst/gstregistrypool.h: remove
* gst/gsttypefind.c:
* gst/gsttypefindfactory.c:
* gst/gsturi.c:
* tools/Makefile.am:
* tools/gst-compprep.c:
* tools/gst-inspect.c:
* tools/gst-register.c: remove
* tools/gst-xmlinspect.c:
Registry rewrite. Changes registry from being a file created
by a tool into a simple cache file created automatically by
libgstreamer. Removed gst-register (because it's no longer
needed). Remove registry pools, because we only have one
registry implementation (XML). Fix up other subsystems as
necessary.
Original commit message from CVS:
* check/generic/states.c: (GST_START_TEST):
Cleanup can be done at the end.
* gst/gsttask.c: (gst_task_get_type), (gst_task_finalize),
(gst_task_func), (gst_task_cleanup_all), (gst_task_set_lock),
(gst_task_get_state), (gst_task_start), (gst_task_pause):
Oh boy.. Thanks for finding this, Thomas.
Original commit message from CVS:
* check/generic/states.c: (GST_START_TEST):
Make sure all tasks are stopped.
* check/gst/gstbin.c: (GST_START_TEST):
Unref after usage for proper valgrinding.
* gst/gstpad.c: (gst_pad_finalize), (gst_pad_stop_task):
Really wait for the task to stop before destroying the
mutex.
* gst/gstqueue.c: (gst_queue_sink_activate_push),
(gst_queue_src_activate_push):
Small cleanups. Don't stop the task when we did not start
it.
* gst/gsttask.c: (gst_task_get_type), (gst_task_init),
(gst_task_func), (gst_task_cleanup_all), (gst_task_set_lock),
(gst_task_get_state), (gst_task_start), (gst_task_pause),
(gst_task_join):
* gst/gsttask.h:
Protect the stream lock with the object lock.
Disallow setting the stream lock when running.
Add cleanup_all to wait for the threadpool to finish.
Remove code to autoallocate a mutex if none was provided.
Add _join() to wait for a task to stop.
Protect the thread pool with a global lock.
Original commit message from CVS:
* check/Makefile.am:
* check/generic/states.c: (GST_START_TEST), (states_suite), (main):
add a test that does a bunch of state changes on elements
needs some fixing for valgrind
* check/states/sinks.c: (gst_object_suite):
whitespace
* gst/gstcaps.h:
add prototype for gst_caps_is_equal_fixed
* gst/gstplugin.c:
* gst/gstregistrypool.c:
doc fixes
