In gst_base_src_start_complete() we do a perform_seek() that will eventually
start the streaming thread which acquires the live lock and then goes to sleep
in the case of appsrc. Right after we perform seek we also try to acquire the
live lock which might then deadlock.
fix this by taking the stream lock before performing the seek. This makes sure
that the streaming thread cannot start and grab the live lock until we are done
and release the stream lock again.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=676048
This allows subclasses to override it, as is necessary for e.g. the
video-crop meta. It is now necessary that after decide_allocation()
there is always a allocator and a configured buffer pool inside the
query.
It causes the timestamp to go wrong, should not cause much of a performance
increase and in the cases where it is faster, it is broken in 0.10 as well.
We should try to review this when rewriting the adapter for 0.11 memory
features.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=674791
... which presently mainly serves to answer SEEKING query negatively
to dissuade upstream encoders from doing any seeking and
"header finalization" (since the returned result of pushing a
sticky event is fairly useless nowadays).
Remove GST_MAJORMINOR and replace it by GST_API_VERSION
Also set GST_VERSION_{MAJOR,MINOR,MICRO,NANO} explicitely
now.
All versions are at 1.0.0 now for the release soon but
API/ABI can still change until the 1.0.0 release.
Next release versions until 1.0.0 will be 0.10.9X and
these will be release candidates. GST_VERSION_* will
nonetheless stay at 1.0.0.0.
This avoids ending up with plenty of pending data (since we'll only
try to parse/push one frame from the incoming buffer).
Fixes increasing memory consumption when parsers aren't linked
Conflicts:
libs/gst/base/gstbaseparse.c
This avoids ending up with plenty of pending data (since we'll only
try to parse/push one frame from the incoming buffer).
Fixes increasing memory consumption when parsers aren't linked
Add an option to control if transform_ip is called in passthrough mode or not.
for elements that don't want to look at the data in passthrough mode, this can
avoid some extra processing, mostly in subclasses.
gst_buffer_take_memory -> gst_buffer_insert_memory because insert is what the
method does.
Make all methods deal with ranges so that we can replace, merge, remove and map
a certain subset of the memory in a buffer. With the new methods we can make
some code nicer and reuse more code. Being able to deal with a subset of the
buffer memory allows us to optimize more cases later (most notably RTP headers
and payload that could be in different memory objects).
Make some more convenient macros that call the more generic range methods.
Add gst_buffer_append() which appends the memory blocks from one buffer to
another. Remove the old inefficient _merge() and _join() methods which forced a
premature memcpy in most cases.
Remove the _is_span() and _span() methods they are not needed anymore now that
we can _append(). Merging and spanning will be delayed until mapping or maybe
not at all when the element can deal with the different memory blocks.
... which controls how to (forcibly) deal with (non-)writable data and
is not necessarily related to identical caps.
In particular, it is also not so helpful anymore with a more advanced
GstVideoFilter subclass which always has a transform_ip method currently,
even though its subclass may not have a corresponding _ip method.
Improve the docs of the get/pull_range functions, define the lifetime of the
buffer in case of errors and short reads.
Make sure the code does what the docs say.
Make it so that one can specify a buffer for get/pull_range where the downstream
element should write into. When passing NULL, upstream should allocate a buffer,
like in 0.10.
We also need to change the probes a little because before the pull probe, there
could already be a buffer passed. This then allows us to use the same PROBE
macro for before and after pulling.
While we're at the probes, make the query probe more powerful by handling the
GST_PAD_PROBE_DROP return value. Returning _DROP from a query probe will now
return TRUE upstream and will not forward the probe to the peer or handler.
Also handle _DROP for get/pull_range properly by not dispatching to the
peer/handler or by generating EOS when the probe returns DROP and no buffer.
Make filesrc handle the non-NULL buffer passed in the get_range function and
skip the allocation in that case, writing directly into the downstream provided
buffer.
Update tests because now we need to make sure to not pass a random value in the
buffer pointer to get/pull_range
Separate the bufferpool and allocator hints in the allocation query, some
of the values don't always make sense together.
Keep the bufferpool and its configuration together.
Keep the allocator and its parameters together.
Allow for multiple bufferpool configurations in the query.
Group the extra allocation parameters in a GstAllocationParams structure to make
it easier to deal with them and so that we can extend them later if needed.
Make gst_buffer_new_allocate() take the GstAllocationParams for added
functionality.
Add boxed type for GstAllocationParams.
Change gst_allocator_alloc() so that we can also spicify flags and padding.
Add 2 new flags to mark the memory 0 prefixed/padded. This allows us to
remove some resizes in the base classes.
When allocating memory, memset prefix and padding with 0 when the flags tell
us to.
On resize, clear the zero padding flags if we can't guarantee the memory is
still 0 filled.
Update tests.
Negotiating padding is needed on second thought so include it in the
ALLOCATION query.
Make the bufferpool take padding into account when allocating.
Make basesrc take padding into account.
Use padding and prefix when allocating in basetransform.
Rework some caps operations so they don't rely on writable caps but instead take
ownership of the input caps and do _make_writable() only when needed.
Remove some const from caps functions, it does not make much sense for
refcounted objects and does not allow us to return a refcount to the const input
caps.
Rework the base classes fixate vmethods to not operate on the caps in-place.
All this saves us around 30% of caps and structure copy and new operations.
Answer the allocation query with FALSE when we are not negotiated yet because at
that point we have no idea if we need to proxy the allocation query or not.
Previously only the last would be pushed, which would cause
invalid running times downstream. This also fixes the handling
of update newsegment events.
The segment start adjustment code in pull mode should never trigger
anymore because the bisection code earlier would have already made
sure that we're at the desired position.
Also move the gap handling some lines below after sending the currently
configured segments. Otherwise we might fill gaps in a segment that is
not configured downstream yet.
Add a vmethod to filter metadata that should be passed upstream. By default,
don't pass anything.
Add a vmethod to transform metadata from the input buffer to the output buffer.
By default, nothing is transformed or copied.
Remove metadata that describes the particular memory of the buffer it is
attached to. We need to do this because in non-passthrough mode we will allocate
new memory for our output buffer.
Improve the propose allocation vmethod by passing the downstream allocation
query to it. This way the vmethod implementation can use properties of the
downstream allocation to generate the upstream query result. If there is no
downstream quety, it means that the element is working in passthrough mode.
Implement a default decide_allocation.
Rename gst_base_transform_suggest to gst_base_transform_reconfigure_sink because
that is what it does. Also remove the caps and size because that is not needed.
Rename gst_base_transform_reconfigure to gst_base_transform_reconfigure_src.
Remove some old unused code in capsfilter.
Always call the propose_allocation method and provide a default implementation
that passes the query on in passthrough mode so that subclasses can also call
this. Also pass if the transform is in passthrough mode so that the
implementation can adjust its algorithm.
Make it possible to also implement non-inplace fixate functions. Let the fixate
function make the caps writable when needed because some fixate functions might
not need to modify the caps.
Make it possible to configure a GDestroyNotify and user_data for
gst_memory_new_wrapped() this allows for more flexible wrapping of foreign
memory blocks.
Use a frame flag to signal to subclass it should reset any retained
state w.r.t. frame parsing since the frame being passed is 'new',
i.e. not related to previously passed and processed data.
Make the pad activation code look more like other activation code.
Only start the sinkpad task when we decide to activate in pull mode, when we
later add srcpad pullmode this will be needed.
... so as to allow the push-mode case to provide data to subclass
on a buffer by buffer basis (as in regular forward case), rather
than all buffers of a fragment chucked together.
Also refactor buffer handling some more, and add some debug.
... since that is already handled by _update_duration, or should not be done
altogether if the duration is determined by non-estimated means.
Fixes#669502.
Clean up the prepare_output_buffer function.
Reuse the input buffer when it is writable and when doing an
in-place but non-passthrough transform.
Move the copy-metadata function call to the prepare_output_buffer default
function. If subclasses implement a custom prepare_output_buffer, they must also
copy the metadata themselves.
Remove a useless memory copy because prepare_output_buffer already did that.
Place the allocator object in the ALLOCATION query instead of the name. This
allows us to exchange allocators that are not in the global pool of allocators.
Update elements for the new api
Add refcounting to the GstAllocator object.
Remove const from functions because the allocator is refcounted now.
Rename the vmethods for consistency
Expose the constructor for GstAllocator and add a destroy notify for the
user_data. This should make it possible to create allocators that are not
registered and shared globally along with the possibility to destroy them
properly.
Update defs with new symbols.
And only forward the allocation query if we're working in
passthrough mode if no ::propose_allocation is implemented.
In place buffer transformations will change the buffer
content and require explicit handling of all metas.
When _clear gets called between _map and _unmap, buffers
will be unreffed. If the adapter was mapped, memory leaks
may occur.
While calling _clear between _map and _unmap does not seem
like such a great idea, this is possible in the audio
encoder base class, as _clear may be called in _finish_frame.
Since the audio encoder relies on flushing to keep track of
timestamps, delaying flushing till after handle_frame seems
dangerous.
So, we unmap on clear, as the next unmap will do nothing.
This makes _clear safe to call between _map and _unmap,
while avoiding leaking the mapped buffer.
https://bugzilla.gnome.org/show_bug.cgi?id=664133
GstBaseParse was not clearing its adapter on reset causing
problems when a pipeline went for example from PLAYING to NULL
state and then back to PLAYING again. The data from the last
stream would be used in the parser.
https://bugzilla.gnome.org/show_bug.cgi?id=667444
There are many good use cases for GstIndex and we want
to add it back again in some form, but possibly not with
the current API, which is very powerful (maybe too powerful),
but also a bit confusing. At the very least we'd need to
make the API bindings-friendly.
Conflicts:
libs/gst/base/gstbasetransform.c
libs/gst/controller/gstinterpolationcontrolsource.c
libs/gst/controller/gstlfocontrolsource.c
plugins/elements/gstfilesrc.c
Dit not merge controller or basetransform changes.
Conflicts:
gst/gstindexfactory.c
libs/gst/base/gstbasetransform.c
plugins/elements/gstfakesink.c
plugins/elements/gstfakesrc.c
plugins/elements/gstidentity.c
plugins/elements/gstinputselector.c
plugins/elements/gstoutputselector.c
Note: did not merge any of the basetransform changes from 0.10.
Add a method to enable async start behaviour. The subclass can then complete the
start operation from any other thread by caling gst_base_src_start_complete().
The base class can wait for the start to complete with
gst_base_src_start_wait().
Add private replacements for deprecated functions such as
g_mutex_new(), g_mutex_free(), g_cond_new() etc., mostly
to avoid the deprecation warnings. We can't change most of
these in 0.10 because they're part of our API and ABI.
Merge the render_object code with the chain method. It is only called from there
and there are quite a few variables that can be reused to makes things less
confusing.
Add new wait_eos vmethod to wait for the eos timeout before posting the EOS
message on the bus.
Add default event handler. Move the default event actions in there. Call the
event vmethod from the pad event handler. Subclasses are now supposed to chain
up to the parent event handler or unref the event and do their own thing.
Avoid passing unused parameters to functions.
Rename the last-buffer property to last-sample and make it return the new
GstSample type so that we can include caps and timing info in one nice bundle.
This makes sure that we get correct and complete caps. The suggested caps
could be incomplete, e.g. video/x-raw-rgb without any fields, and by
intersecting with the peer caps we get something usable.
Fixes bug #662199.
Using gst_pad_proxy_get_caps() breaks backwards compatibility with old
parsers because it will propagate the other side's fields like "parsed"
and "framed" and also breaks parser/converters.
Fixes bug #664221.
In 0.10, this can be done with a one-liner by using GST_BUFFER_DATA/SIZE with
put_data. A 0.11 user has to resort to gst_buffer_map, which is less convenient
and might require a memcpy internally.
API: gst_byte_writer_put_buffer()
Fixes compiler warnings on OSX:
gstadapter.h:82: warning: type qualifiers ignored on function return type
gstadapter.c:412: warning: type qualifiers ignored on function return type
const gpointer is not the same as gconstpointer or const void *.
https://bugzilla.gnome.org/show_bug.cgi?id=664491
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.
Turns some boolean arguments in the scheduling query to flags, which are easier
to extend and makes the code easier to read.
Make extra methods for configuring and querying the supported scheduling modes.
This should make it easier to add new modes later.
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.
Don't rely on the return value of a vmethod to trigger the default
implementation but make a real defaul implementation of the method that the
subclass can chain up to.
Make a default implementation of the transform_caps vmethod so that subclasses
can call into it.
Make a default implementation of transform_size.
Avoid doing something in the vmethod trampoline.
The fixate caps function was not used externally and we have vmethods in the
base classes where it is needed.
Update some docs.
simplify some fixate functions in the base classes. Also pass the untruncated
caps to the vmethod.
... in the extent that a non-waiting pad (so indicated by newsegment)
turns out to provide the best buffer, which is then forced to waiting
for book-keeping purposes, but that should only be temporary.
See bug #415754.
This handles muxing of sparse/subtitle streams and has
lots of cleanup. Still missing is special support for
live streams but this can be added later without breaking
API/ABI.
Based on the version from the videomixer plugin.
https://bugzilla.gnome.org/show_bug.cgi?id=415754
Speeds up negotiation a fair bit on a contrived pipeline
with a dozen colorspace conversions.
Hopefully clears out the cache every time it ought to.
https://bugzilla.gnome.org/show_bug.cgi?id=662291
API: GstBaseParseClass::detect()
This is called with the first buffers until the subclass has finished detection
and only afterwards the original buffers are handled as before. The vfunc allows
detection of the stream format without breaking the upstream framing.
Adds a getcaps function to the sink pad to make parsers propagate
downstream caps restrictions to upstream.
The pipeline "audiotestsrc num-buffers=100 ! faac ! aacparse !
"audio/mpeg, version=(int)4, stream-format=(string)adts" ! filesink"
wouldn't work because aacparse wouldn't propagate the adts restriction
upstream to faac.
This patch adds a default getcaps to the sink pad to simply proxy
downstream caps and also adds a 'get_sink_caps' function pointer
to GstBaseParseClass for subclasses that need more refined getcaps.
https://bugzilla.gnome.org/show_bug.cgi?id=661874
If a class extending basesrc doesn't set blocksize, basesrc
would try to allocate a (guint)-1 sized buffer, which is enormous
and likely would fail.
Avoid it and error out.
There's no code that uses it other than multiqueue, so make it private
to multiqueue for now. That way we can also do optimisations that
require API/ABI breaks. If anyone ever wants to use it, we can still
make it public again.
Some elements (such as videorate) might push buffers early,
for instance in in transform_ip. We want events (and in particular
any NEWSEGMENT event) to be pushed before that.
This fixes transmageddon wedging on converting a file starting
with a non zero offset to Ogg.
https://bugzilla.gnome.org/show_bug.cgi?id=660165
Otherwise elements like capsfilter will return ANY caps if no
peer is present instead of the filter caps. The transform_caps()
vfunc could do transformations to the template caps that do not
result in the unmodified template caps.
Name the allocation vmethod on srcpad decide_allocation because source pads will
have to decide what allocation parameters will be used.
Name the allocation vmethod on sinkpads propose_allocation because they will
need to configure the allocation query with a proposed values for upstream.
Wim suggested that using GstPadDirection instead of a GstPad in the
arguments to the new query vfunc would be more consistent with the other
functions.
Remove the negotiation from the state change function, it causes data transfer
and bufferpool negotiation, which is not supposed to be done. Since we have the
reconfigure state on the pad, the create function will do the negotiation as
soon as it gets in the streaming thread.
Don't change the state of the bufferpool when going between PAUSED and PLAYING,
it will dealloc and realloc all buffers, which is clearly too invasive. We will
need to add some other way of unblocking the bufferpool.
Add a vmethod to handle the pad query.
Install a default handler for the pad query.
Add a vmethod to setup the allocation properties.
Use the new query function in filesink
Implement the sink event handling like the src event handler. Make the default
implementation parse and forward the event. This makes it possible to actually
return an error value from the event handler.
Remove the requirement to have to return a ref to the input buffer when in
passthrough mode. This saves a few ref/unref cycles and fixes another 0.11
FIXME.
Make a new copy_metadata vmethod and move the code to copy the timestamps, flags
and offsets into a default implementation. This will allow us to give the
subclasses a chance to override the copy method.
Move the code for prepare_output_buffer to a default implementation. this allows
us to simplify some things and have subclasses call into the default
implementation when needed.
Remove the caps and size from the prepare_output_buffer function. with
bufferpools and capsnego done differently, we don't need this in most cases and
if we do, we can simply use the transform_size function and get the caps from
the srcpad.
Don't mix messages and pads and tags.
Make the sink post tag messages when a tag event is received.
Since tags are sticky on pads now, they can be retrieved from there
when needed.
Add an index to gst_buffer_take_memory() so that we can also insert memory at a
certain offset. This is mostly interesting to prepend a header memory block to
the buffer.
Make a new method to allocate a buffer + memory that takes the allocator and the
alignment as parameters. Provide a macro for the old method but prefer to use
the new method to encourage plugins to negotiate the allocator properly.
Add a new fill virtual method to basesrc. The purpose of this method is to fill
a provided buffer with data.
Add a default implementation of the create method that allocates a buffer and
calls the fill method on it. This would allow the base class to implement
bufferpool and allocator negotiation on behalf of the subclasses.
Fix the blocksize property.
Make filesrc use the new fill method.
Add a boolean to the flush_stop event to make it possible to implement flushes
that don't reset_time.
Make basesink post async_done with the reset_time property from the flush stop
event.
Fix some unit tests
This allows subclass to indicate that size reported by src may not be static
and should as such be updated regularly, rather than only when really
needed.
Particular examples are filesrc or fdsrc reading from a file that is still
growing (e.g. being downloaded).
Fixes#652037.
This reverts commit 934faf163c.
Original commit leads to possibly sending newsegment event downstream
in pull mode. In push mode, quite some downstream elements
are likely to only expect newsegment event following a seek they performed
and as such may have their state messed up.
Move the flag to indicate that a new_base_time should be distributed to the
pipeline, from the async_start to the async_done message. This would allow us to
decide when to reset the pipeline time based on other reasons than the
FLUSH_START event.
The main goal eventually is to make the FLUSH events not reset time at all but
reset the time based on the first buffer or segment that prerolls the pipeline
again.
Instead of passing it structure by structure. This allows
better optimized transform_caps functions and allows better
transformation decisions.
See bug #619844.
Don't error out when the allocation query returns success.
Do bufferpool query after we pushed the caps event downstream so that we can get
a good bufferpool suggestion.
Also proxy the bufferpool query downstream when we operate in in_place mode.
Avoid installing a setcaps function on the srcpad and calling the setcaps
function, we can do more efficiently with sending the event ourself and calling
our vmethod.
