When releasing the request pad, first remove it from the element and then
deactivate it. If we do it the other way around, a gst_pad_push on the element
might return wrong-state before we had a chance to detect the removed pad in the
chain function.
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.
1) We need to lock and get a strong ref to the parent, if still there.
2) If it has gone away, we need to handle that gracefully.
This is necessary in order to safely modify a running pipeline. Has been
observed when a streaming thread is doing a buffer_alloc() while an
application thread sends an event on a pad further downstream, and from
within a pad probe (holding STREAM_LOCK) carries out the pipeline plumbing
while the streaming thread has its buffer_alloc() in progress.
Make sure property names passed to g_object_notify() are in the canonical form
(ie. "last-message" not "last_message"), so that g_param_spec_pool_lookup()
doesn't have to do strdup/canonicalize/free for every single notify call.
This only applies when building against older GLib versions (< 2.26).
Protect the ->removed field with the object lock as well. Take the DYN lock
earlier so that we can mark the pad removed and avoid a race in pad_alloc.
Fixes#606435
It is not easy to setup a tee on the fly, thus apps need to add them always if
they might need them. This changes the code so, that if only one src-pad is
active, we push buffers directly. In the normal code path all buffers are pushed
with an extra ref, that forces followup inplace elements to copy the data.
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.
Original commit message from CVS:
* plugins/elements/gsttee.c: (gst_tee_handle_buffer):
Check for changed pads-list before checking the last returned
GstFlowReturn because the pad could have been removed and we
need to ignore the value in that case.
Original commit message from CVS:
* plugins/elements/gsttee.c: (gst_tee_handle_buffer):
Fix flow aggregation of tee. Error out immediately for all flow returns
except OK and NOT_LINKED, return NOT_LINKED if all pads are not linked
and return OK if at least one pad is linked.
Before we errored out on "fatal" flow returns (i.e. not for WRONG_STATE)
and otherwise returned the flow return of the last pad, which is wrong.
* tests/check/elements/tee.c: (_fake_chain), (_fake_chain_error),
(GST_START_TEST), (tee_suite):
Add unit tests for the flow aggregation.
Original commit message from CVS:
Patch by: Ole André Vadla Ravnås <ole.andre.ravnas at tandberg com>
* plugins/elements/gsttee.c: (gst_tee_finalize), (gst_tee_init),
(gst_tee_request_new_pad), (gst_tee_release_pad),
(gst_tee_find_buffer_alloc), (gst_tee_buffer_alloc):
* plugins/elements/gsttee.h:
Protect pad_alloc with a new lock so that we can be sure that nothing is
performing a pad_alloc when removing the pad. Fixes#547835.
* tests/check/elements/tee.c: (buffer_alloc_harness_setup),
(buffer_alloc_harness_teardown), (app_thread_func),
(final_sinkpad_bufferalloc), (GST_START_TEST), (tee_suite):
Added testcase for shutdown race.
Original commit message from CVS:
* plugins/elements/gsttee.c:
* plugins/elements/gsttee.h:
Reverting as not everything is clear yet. Needs some general design
work.
Original commit message from CVS:
* plugins/elements/gsttee.c: (gst_tee_base_init),
(gst_tee_request_new_pad), (gst_tee_release_pad),
(gst_tee_find_buffer_alloc), (gst_tee_buffer_alloc),
(gst_tee_do_push), (clear_pads), (gst_tee_handle_buffer),
(gst_tee_chain):
Be a lot smarter when deciding what srcpad to use for proxying
the buffer_alloc. Also handle pad added/removed when doing so.
Fixes#357959.
Keep track of what pads we already pushed on in case we have pads
added/removed while pushing. Fixes#374639
* tests/check/Makefile.am:
* tests/check/elements/tee.c: (handoff), (GST_START_TEST),
(tee_suite):
Added unit test for pad resync.
Original commit message from CVS:
2007-02-09 Andy Wingo <wingo@pobox.com>
* tests/check/pipelines/simple-launch-lines.c (test_tee): Add
tests for push and pull tee behavior.
* plugins/elements/gsttee.h:
* plugins/elements/gsttee.c: Describe has-sink-loop better, and
mark as deprecated as well as unimplemented. It was a crack idea.
Add support for tee operating in pull mode, off by default.
Original commit message from CVS:
* docs/plugins/gstreamer-plugins.args:
* docs/plugins/inspect/plugin-coreelements.xml:
* docs/plugins/inspect/plugin-coreindexers.xml:
Update element args.
* gst/gstsystemclock.c:
Small comment update.
* plugins/elements/gsttee.c: (gst_tee_class_init), (gst_tee_init),
(gst_tee_request_new_pad), (gst_tee_release_pad),
(gst_tee_buffer_alloc), (gst_tee_sink_activate_push),
(gst_tee_sink_activate_pull):
* plugins/elements/gsttee.h:
Some tee loving:
Add default property defines.
Implement release pad function.
Give properties better blubs etc.
Activate pads before adding them to a running tee.
Do simple buffer_alloc on the first requested pad.
Post error when activation fails.
Original commit message from CVS:
* plugins/elements/gsttee.c: (gst_tee_do_push),
(gst_tee_handle_buffer):
Always prefer an actual return value from a src
pad in place of NOT_LINKED. This means we return
WRONG_STATE when all src pads are WRONG_STATE
instead of NOT_LINKED.
Lock when replacing the last message to prevent
racing with the get_property method.
Add debug output
Original commit message from CVS:
2006-01-26 Julien MOUTTE <julien@moutte.net>
* plugins/elements/gsttee.c: (gst_tee_do_push),
(gst_tee_handle_buffer): Apply patch from #328715. Tee now
handles pad being NOT_LINKED or in WRONG_STATE.
Original commit message from CVS:
2005-11-21 Andy Wingo <wingo@pobox.com>
* *.h:
* *.c: Ran scripts/update-macros. Oh yes.
* gst/gstobject.h (GST_OBJECT_GET_LOCK, GST_OBJECT_LOCK)
(GST_OBJECT_TRYLOCK, GST_OBJECT_UNLOCK): Renamed from
GST_GET_LOCK, etc.
* scripts/update-macros: New script. Run it on your files to
change GST_LOCK to GST_OBJECT_LOCK, and the same for UNLOCK as
well.
Original commit message from CVS:
* gst/gstpad.c: (gst_pad_new_from_static_template):
* gst/gstpad.h:
add gst_pad_new_from_static_template functions
* gst/check/gstcheck.c: (gst_check_setup_src_pad),
(gst_check_setup_sink_pad):
* gst/elements/gsttee.c: (gst_tee_init):
and use them
Original commit message from CVS:
* gst/elements/gsttee.c: (gst_tee_do_push), (gst_tee_handle_buffer):
Let's not iterate pads we're not interested in, it avoids getting
sky-high refcounts on sinkpad.
Original commit message from CVS:
* gst/elements/gsttee.c: (gst_tee_init), (gst_tee_do_push):
* gst/gstutils.c: (intersect_caps_func), (gst_pad_proxy_getcaps),
(link_fold_func), (gst_pad_proxy_setcaps):
Leak fixes, the fold functions need to unref the passed object and
_get_parent_*() returns ref to parent.
Original commit message from CVS:
2005-08-28 Andy Wingo <wingo@pobox.com>
* gst/gstutils.h (GST_BOILERPLATE_FULL): Prototype instance_init
as having two arguments instead of just one. Allows superclasses
to access information on subclasses -- see the terrible for() loop
in gtype.c:g_type_create_instance for the reason why. All callers
changed.
Original commit message from CVS:
2005-06-27 Andy Wingo <wingo@pobox.com>
* gst/base/gsttypefindhelper.c (gst_type_find_helper): Unref any
remaining buffer.
* gst/gsttrace.c (gst_alloc_trace_list_sorted): New helper,
returns a sorted copy of the trace list.
(gst_alloc_trace_print_live): New API, only prints traces with
live objects. Sort the list.
(gst_alloc_trace_print_all): Sort the list.
(gst_alloc_trace_print): Align columns.
* gst/elements/gstttypefindelement.c:
* gst/elements/gsttee.c:
* gst/base/gstbasesrc.c:
* gst/base/gstbasesink.c:
* gst/base/gstbasetransform.c:
* gst/gstqueue.c: Adapt for pad activation changes.
* gst/gstpipeline.c (gst_pipeline_init): Unref after parenting
sched.
(gst_pipeline_dispose): Drop ref on sched.
* gst/gstpad.c (gst_pad_init): Set the default activate func.
(gst_pad_activate_default): Push mode by default.
(pre_activate_switch, post_activate_switch): New stubs, things to
do before and after switching activation modes on pads.
(gst_pad_set_active): Take a boolean and not a mode, dispatch to
the pad's activate function to choose which mode to activate.
Shortcut on deactivation and call the right function directly.
(gst_pad_activate_pull): New API, (de)activates a pad in pull
mode.
(gst_pad_activate_push): New API, same for push mode.
(gst_pad_set_activate_function)
(gst_pad_set_activatepull_function)
(gst_pad_set_activatepush_function): Setters for new API.
* gst/gstminiobject.c (gst_mini_object_new, gst_mini_object_free):
Trace all miniobjects.
(gst_mini_object_make_writable): Unref the arg if we copy, like
gst_caps_make_writable.
* gst/gstmessage.c (_gst_message_initialize): No trace init.
* gst/gstghostpad.c (gst_proxy_pad_do_activate)
(gst_proxy_pad_do_activatepull, gst_proxy_pad_do_activatepush):
Adapt for new pad API.
* gst/gstevent.c (_gst_event_initialize): Don't initialize trace.
* gst/gstelement.h:
* gst/gstelement.c (gst_element_iterate_src_pads)
(gst_element_iterate_sink_pads): New API functions.
* gst/gstelement.c (iterator_fold_with_resync): New utility,
should fold into gstiterator.c in some form.
(gst_element_pads_activate): Simplified via use of fold and
delegation of decisions to gstpad->activate.
* gst/gstbus.c (gst_bus_source_finalize): Set the bus to NULL,
help in debugging.
* gst/gstbuffer.c (_gst_buffer_initialize): Ref the buffer type
class once in init, like gstmessage. Didn't run into this issue
but it seems correct. Don't initialize a trace, gstminiobject does
that.
* check/pipelines/simple_launch_lines.c (test_stop_from_app): New
test, runs fakesrc ! fakesink, stopping on ::handoff via a message
to the bus.
(assert_live_count): New util function, uses alloc traces to check
cleanup.
* check/gst/gstghostpad.c (test_ghost_pads): More refcount checks.
To be modified when unlink drops the internal pad.
Original commit message from CVS:
* gst/elements/gsttee.c: (gst_tee_chain), (gst_tee_loop),
(gst_tee_sink_activate):
Core takes STREAM_LOCK for us now.
Original commit message from CVS:
Next big merge.
Added GstBus for mainloop integration.
Added GstMessage for sending notifications on the bus.
Added GstTask as an abstraction for pipeline entry points.
Removed GstThread.
Removed Schedulers.
Simplified GstQueue for multithreaded core.
Made _link threadsafe, removed old capsnego.
Added STREAM_LOCK and PREROLL_LOCK in GstPad.
Added pad blocking functions.
Reworked scheduling functions in GstPad to prepare for
scheduling updates soon.
Moved events out of data stream.
Simplified GstEvent types.
Added return values to push/pull.
Removed clocking from GstElement.
Added prototypes for state change function for next merge.
Removed iterate from bins and state change management.
Fixed some elements, disabled others for now.
Fixed -inspect and -launch.
Added check for GstBus.
Original commit message from CVS:
Remove threadsafe properties. Fix elements because GObject
complains when installing a property before declaring a
set/get_property handler.
Rearrange gstelement.h file, use STATE macros for state locks.
Free mutexes in the finalize method instead of dispose.
Original commit message from CVS:
First THREADED backport attempt, focusing on adding locks and
making sure the API is threadsafe. Needs more work. More docs
follow this week.
Original commit message from CVS:
- removing some trailing commas in enums to please C90 only compilers (AIX, Forte, ...)
('should' fix#143290)
should/could we compile gst with -ansi -pedantic to catch those in the build-bots ?
Original commit message from CVS:
2004-01-25 Benjamin Otte <in7y118@public.uni-hamburg.de>
* gst/elements/gstbufferstore.c: (gst_buffer_store_get_buffer):
fix bug where buffer was not assembled correctly
* gst/elements/gsttee.c: (gst_tee_class_init), (gst_tee_init):
silence by default
* gst/elements/gsttypefind.c: (gst_type_find_element_chain):
only seek if there's no more buffers that could work without seeking
Original commit message from CVS:
* gst/elements/gsttee.c: (gst_tee_init), (gst_tee_request_new_pad):
Remove usage of gst_pad_proxy_fixate.
* gst/gstcaps.c: (gst_caps_append), (gst_caps_append_structure),
(gst_caps_split_one), (gst_caps_replace):
Add poisoning code.
* gst/gstmarshal.list:
Add pointer__pointer for fixate signal
* gst/gstpad.c: (gst_real_pad_class_init),
(_gst_real_pad_fixate_accumulator), (gst_pad_link_fixate),
(_gst_pad_default_fixate_func), (gst_pad_proxy_fixate),
(gst_pad_set_explicit_caps), (gst_pad_template_new):
Add poisoning code. Add fixate signal on RealPad. Change
set_explicit_caps() to take const GstCaps, like try_set_caps().
* gst/gstpad.h:
* testsuite/caps/Makefile.am:
* testsuite/caps/app_fixate.c: Add a test for the fixate signal
Original commit message from CVS:
* gst/elements/gstaggregator.c:
* gst/elements/gsttee.c:
Use gst_pad_proxy_getcaps().
* gst/gstpad.c:
* gst/gstpad.h:
Add gst_pad_proxy_getcaps(), which filter elements can use
as a generic getcaps implementation.
Fix gst_pad_get_allowed_caps(). It just wasn't doing what
was advertised.
Original commit message from CVS:
s/GstBuffer/GstData/ in the API where you can pass events. Fix the plugins to deal with that. Fixes#113488. Also includes scheduler patches, and probably fixes some queue bugs relating to events and buffers.
Original commit message from CVS:
GST_DEBUG reorganization
This is a big diff (ca 450k), containing loads of stuff:
- gstinfo.[ch] complete rewrite
- changing of all GST_DEBUG messages to reflect that change
- reorganization of subsystem disabling
- addition of gstconfig.h.in so we can track the disablings
- <gst/gst.h> does not include <unistd.h> and <config.h> anymore
- documentation updated for gstinfo stuff (build the docs yourself to know what changed)
- bugfixes for making of the docs (files from CVS are not deleted anymore
- testsuite for debugging changes in testsuite/debug
expect breakage
Original commit message from CVS:
- the result from the getcaps function should be sunken (gst_caps_sink)
instead of unreffed. Not sure that's the right thing to do...
Original commit message from CVS:
Convert %lld and %llu in printf formats to G_G[U]INT64_FORMAT. Fix
pointer<->int conversion. Fixes warnings on alpha.
Original commit message from CVS:
- Make the return GList of gst_element_get_pad_list as const
- Pad scheduler refactoring
- Remove gst_pad_set/unset_scheduler as it's not needed
- Reimplement gst_pad_get_scheduler using the parent scheduler
- Remove gst_pad_peek as it cannot work reliably
- Remove the pad bufpen, replace with scheduler private gpointer
- Make queue use the new _get_scheduler implementation
- Remove _pad_unset_scheduler from GstScheduler
- Remove LOOP_SEEK mode from aggregator
- Other fixes for API changes.
Original commit message from CVS:
This changes an important part of the plugin API, gst_pad_try_set_caps() no longer returns a boolean, it now returns a GstPadConnectReturn, which makes much more sense than a boolean. All plugins have also been changed, so don't worry ;)
Original commit message from CVS:
* a hack to work around intltool's brokenness
* a current check for mpeg2dec
* details->klass reorganizations
* an element browser that uses details->klass
* separated cdxa parse out from the avi directory
Original commit message from CVS:
commit to make gstreamer follow the gtk function/macro naming conventions:
GstPadTemplate <-> gst_pad_template <-> GST_PAD_TEMPLATE
and the same for *factory and typefind.
Original commit message from CVS:
- Remove the propsprivate header file
- Added new API for properties.
- Moved the clock distribution to the scheduler.
- Removed the lock from GstCaps
- Added boxed types for Caps/Props
- Simplified the clock, new systemclock implementation
- Removed deprecated element_info/send_event functions
- First step at exposing more info in the pad_connect functions
- Queue cleanup
- Make the scheduler aware of other schedulers inside it
- Added the _SELF_SCHEDULABLE flag to gstthread
- Removed _get_widget from _utils, changed to new props API
- Make fakesink sync on timestamps when requested
- Removed the offset notify from filesrc
- Added a fast scheduler
- some scheduler cleanups.
