We reset all the waiting streams, let them push another buffer to
see if they're now active again. This allows faster switching
between streams and prevents deadlocks if downstream does any
waiting too.
Also improve locking a bit, srcresult must be protected by the
multiqueue lock too because it's used/set from random threads.
Add the pad mode to the activate function so that we can reuse the same function
for all activation modes. This makes the core logic smaller and allows for some
elements to make their activation code easier. It would allow us to add more
scheduling modes later without having to add more activate functions.
Remove the getcaps function on the pad and use the CAPS query for
the same effect.
Add PROXY_CAPS to the pad flags. This instructs the default caps event and query
handlers to pass on the CAPS related queries and events. This simplifies a lot
of elements that passtrough caps negotiation.
Make two utility functions to proxy caps queries and aggregate the result. Needs
to use the pad forward function instead later.
Make the _query_peer_ utility functions use the gst_pad_peer_query() function to
make sure the probes are emited properly.
See #651514 for details. It's apparently impossible to write code
that avoids both type punning warnings with old g_atomic headers and
assertions in the new. Thus, macros and a version check.
This reverts commit cf4fbc005c.
This change did not improve the situation for bindings because
queries are usually created, then directly passed to a function
and not stored elsewhere, and the writability problem with
miniobjects usually happens with buffers or caps instead.
Improve GstSegment, rename some fields. The idea is to have the GstSegment
structure represent the timing structure of the buffers as they are generated by
the source or demuxer element.
gst_segment_set_seek() -> gst_segment_do_seek()
Rename the NEWSEGMENT event to SEGMENT.
Make parsing of the SEGMENT event into a GstSegment structure.
Pass a GstSegment structure when making a new SEGMENT event. This allows us to
pass the timing info directly to the next element. No accumulation is needed in
the receiving element, all the info is inside the element.
Remove gst_segment_set_newsegment(): This function as used to accumulate
segments received from upstream, which is now not needed anymore because the
segment event contains the complete timing information.
This reverts commit 9ef1346b1f.
Way to much for one commit and I'm not sure we want to get rid of the pad caps
just like that. It's nice to have the buffer and its type in onw nice bundle
without having to drag the complete context with it.
Remove pad_alloc and all references. This can now be done more efficiently and
more flexible with the ALLOCATION query and the bufferpool objects. There is no
reverse negotiation yet but that will be done with an event later.
Use the string provided by the caller for the sinkpad name
if possible. Note that all sanity checking for this name
is already done in GstElement.
Fixes Bug #645931
Instead return after every iteration, which makes sure that the
stream lock is released for a short time after every iteration,
task state changes are checked, etc and this allows the task
to be stopped properly.
When a downstream element returns GST_FLOW_UNEXPECTED we want to:
* let the dataqueue task running
* forward the flow return upstream.
This allows upstream elements to push EOS, and have that EOS event come
downstream.
Fixes#609274
When we receive an UNEXPECTED flowreturn from downstream, we must not shutdown
the pushing thread because upstream will at some point push an EOS that we still
need to push further downstream.
To achieve this, convert the UNEXPECTED return value to OK. Add a fixme so that
we implement the right logic to propagate the flowreturn upstream at some point.
Also clean up the unit test a little.
Fixes#608136
There's not much point in using GST_DEBUG_FUNCPTR with GObject
virtual functions such as get_property, set_propery, finalize and
dispose, since they'll never be used by anyone anyway. Saves a
few bytes and possibly a tenth of a polar bear.
Add support for buffering mode where we post BUFFERING messages based on the
level of the queues. It currently operates on the first queue that goes over or
under the high/low thresholds.
In buffering mode we want to ignore the max visible items to decide when the
queue is filled. Instead, we only look at the number of bytes and/or time in the
queue.
Don't shadow the sq argument in the underrun_cb function but use
a different variable name to iterate the other queues.
Use the same variable name in the overrun_cb function.
We know earlier on in the code whether we're handling an event or a buffer,
just pass that information through.
This commit and the previous commit reduce instruction fetch:
* when pushing buffer (_chain) by 10%
* when popping buffer (_loop) by 3%
The task will always exist as long as its owner (i.e. the pad) and that
owner's owner (i.e. multiqueue) exist.
Reduces the number of instruction fetches by 36%.
Pads have their GstSingleQueue stored as element private data
so there's no need to iterate over the list of single queues
every time. Also every pad only has a single internal link so
use a single iterator instead of a complex custom iterator.
Set the element private data of the pad to NULL when freeing the
single queue.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c:
* plugins/elements/gstmultiqueue.h:
revert extra-size-buffers stuff, caused some race conditions
and extra-size-buffers is not used anymore. Docs needs some updates
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c:
Don't update the cur_time on GST_CLOCK_TIME_NONE (#537804) and don't
activate the pads if they are added in STATE_NULL.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c:
Add documentation for the signals to push our core plugin docs
coverage back up to 100%.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (single_queue_overrun_cb),
(single_queue_underrun_cb):
When trying to make room in the queue, bump the max allowed buffers
bigger than the current amount of buffers in the queue. this fixes some
nasty deadlocks in multiqueue when dynamically changing the limits of
the queue.
Original commit message from CVS:
* docs/design/part-synchronisation.txt:
Update some docs.
* docs/plugins/Makefile.am:
* docs/plugins/gstreamer-plugins-docs.sgml:
* docs/plugins/gstreamer-plugins-sections.txt:
* plugins/elements/gstmultiqueue.c:
Add multiqueue to the docs.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_set_property),
(gst_multi_queue_request_new_pad), (gst_single_queue_flush),
(gst_multi_queue_loop), (gst_multi_queue_sink_activate_push):
Make it so that pads are considered linked until a buffer is pushed
and discovered otherwise. This avoids problems with decodebin2 hanging
after a seek in the filesrc ! decodebin2 name=d ! fakesink d. ! fakesink
case.
Make sure we lock the multiqueue when updating the max-size properties.
Fix a crash on Solaris in a debug statement in get_request_pad that
passes a NULL string to GST_DEBUG.
* tests/check/elements/multiqueue.c: (mq_dummypad_chain),
(run_output_order_test):
Fix the test to allow the first buffer on not-linked pads to come out
of sequence while multiqueue discovers that they are not-linked.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_single_queue_push_one),
(gst_single_queue_new):
* plugins/elements/gstqueue.c: (gst_queue_init),
(gst_queue_push_one):
Fix queue negotiation. If acceptcaps unconditionally returns TRUE,
upstream is tricked into thinking it can suggest a format downstream
while downstream does not support that format. The real problem is that
core calls acceptcaps when pushing a buffer with new caps, for which we
do a little workaround by setting the caps on the srcpad ourselves
before pushing the buffer (until this is figured out). Fixes#486758.
Original commit message from CVS:
Patch by: Mark Nauwelaerts <manauw at skynet be>
* plugins/elements/gstmultiqueue.c:
(gst_multi_queue_get_internal_links), (apply_buffer),
(single_queue_overrun_cb), (gst_single_queue_new):
Implement non-default GstPadIntLinkFunction for multiqueue pads so that
the pipeline layout can be tracked correctly. Fixes#453732.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (apply_buffer),
(single_queue_overrun_cb):
When figuring out when a queue is filled, use our internal time estimate
based on segments, just like check_full does.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_init),
(gst_single_queue_flush), (apply_segment), (apply_buffer),
(gst_single_queue_push_one), (gst_multi_queue_loop),
(gst_multi_queue_sink_activate_push), (gst_multi_queue_sink_event),
(gst_multi_queue_src_activate_push), (wake_up_next_non_linked),
(compute_high_id), (gst_single_queue_new):
* plugins/elements/gstmultiqueue.h:
Take the multiqueue lock when updating the fill level so we don't get
confused.
After applying a buffer or event on the src pad segment, make sure to
call gst_data_queue_limits_changed() to get the data queue to unblock
and check the filled state again.
Rework the not-linked pad handling so the logic is that not-linked
pads can push as fast as they like, but only so they never get
ahead of any linked pads.
* tests/check/elements/multiqueue.c: (mq_sinkpad_to_srcpad),
(mq_dummypad_getcaps), (mq_dummypad_chain), (mq_dummypad_event),
(run_output_order_test), (GST_START_TEST), (multiqueue_suite):
Add a test to check that not-linked pads always stay behind
linked pads.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (apply_buffer),
(gst_single_queue_push_one), (gst_multi_queue_chain),
(gst_multi_queue_sink_event):
Make sure we don't reference the buffer/event after we have given away
ownership in the queue.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_single_queue_flush),
(gst_multi_queue_chain), (gst_multi_queue_sink_event):
Update queue state _after_ adding the item in the queue because else we
could end up being full without the element added yet.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_item_destroy),
(gst_multi_queue_item_new):
Don't use GSlice because we don't depend on >= 2.10 yet.
Original commit message from CVS:
* libs/gst/base/gstdataqueue.c: (gst_data_queue_cleanup),
(gst_data_queue_finalize), (gst_data_queue_locked_is_empty),
(gst_data_queue_set_flushing), (gst_data_queue_push),
(gst_data_queue_pop), (gst_data_queue_drop_head),
(gst_data_queue_limits_changed), (gst_data_queue_get_level):
* libs/gst/base/gstdataqueue.h:
Various cleanups.
Added methods to get the current levels and to inform the queue that the
'full' limits changed.
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_init),
(gst_multi_queue_finalize), (gst_multi_queue_set_property),
(gst_single_queue_flush), (update_time_level), (apply_segment),
(apply_buffer), (gst_single_queue_push_one),
(gst_multi_queue_item_steal_object),
(gst_multi_queue_item_destroy), (gst_multi_queue_item_new),
(gst_multi_queue_loop), (gst_multi_queue_chain),
(gst_multi_queue_sink_activate_push), (gst_multi_queue_sink_event),
(gst_multi_queue_getcaps), (gst_multi_queue_src_activate_push),
(gst_multi_queue_src_query), (single_queue_overrun_cb),
(single_queue_underrun_cb), (single_queue_check_full),
(gst_single_queue_new):
Keep track of time in the queue by measuring the difference between
running_time on input and output. This gives more accurate results and
can compensate for segments correctly.
Make a queue by default only 5 buffers deep. We will now increase the
buffer size depending on the filledness of the other queues.
Factor out commong flush code.
Make sure we don't add additional refcounts to buffers when we can avoid
it.
Propagate GstFlowReturn differently.
Use GSlice for intermediate GstMultiQueueItems.
Keep track of EOS.
Resize queues on over and underruns based on filled level of other
queues.
When checking if the queue is filled, prefer to measure in time if we
can and fall back to bytes when no time is known.
* plugins/elements/gstqueue.c:
Fix return value.
Original commit message from CVS:
* libs/gst/base/gstdataqueue.c:
* libs/gst/base/gstdataqueue.h:
* plugins/elements/gstmultiqueue.c: (gst_single_queue_push_one),
(gst_multi_queue_item_new), (gst_multi_queue_chain),
(gst_multi_queue_sink_event):
* tests/check/elements/multiqueue.c: (multiqueue_suite):
Fix multiqueue leaking buffers and events when downstream or the
queue are flushing. Make refcounting assumptions explicit and
document them (shouldn't break existing code that uses it other than
maybe leak miniobjects, but that already happens anyway). Add unit
test for the most common flushing case. Fixes#423700.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_release_pad):
Don't deadlock when releasing a pad - gst_pad_set_active may try
and take the multiqueue lock too.
Original commit message from CVS:
Patch by: Tommi Myöhänen <ext-tommi dot myohanen at nokia dot com>
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_release_pad):
Deactivate pads and free GstSingleQueue with gst_single_queue_free()
when releasing sink pad. Fixes#425400.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_single_queue_free):
Don't leak GCond.
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/multiqueue.c: (setup_multiqueue),
(GST_START_TEST), (multiqueue_suite):
Add some dead simple unit tests for the 'multiqueue' element
(some bits don't work yet and are disabled for now).
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_multi_queue_finalize),
(gst_single_queue_free):
Fix small leak (free GstSingleQueue structure too, not only contents).
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c:
(gst_multi_queue_loop):
Small fix.
(single_queue_overrun_cb), (single_queue_underrun_cb),
(single_queue_check_full), (gst_single_queue_new):
Implement single queue growth system.
This uses the extra-size properties, and will grow single queues by
that much if one goes full whereas there are others empty. This is
called extra-mode in the code.
When a single queue's levels go back below the initial max-size
limits, it is no longer in extra-mode. This is to ensure we don't
consume too much memory.
Fixes#399875
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c:
(gst_multi_queue_src_activate_push), (gst_single_queue_new):
Remove previous hack of unsetting the flushing flag for the source pad
instead of activating it. Instead, fix the source pad activate function
so that it no longer depends on having a parent set or not.
Original commit message from CVS:
* plugins/elements/gstmultiqueue.c: (gst_single_queue_new):
When dynamically creating single queues, activate sinkpad before adding
it.
We should be doing the same thing for the source pad, but we can't
since it would call a method which needs the parent to be set in order
to work propertly. Instead of activating the source pad, we just unset
the flushing flag, which is the minimal requirement for adding a pad
to an element in a state greater than READY.
