The AV1 support multi spatial layers within one TU with different
resolutions, and only the highest spatial layer need to be output.
For example, there are two spatial layer, base level is 800x600
and higher level is 1920x1080. We need to decode both because the
higher level needs base layer as reference, but we only need to output
1920x1080 frames here.
The current manner always renegotiates the caps once we detect the
current picture resolution changes, so we renegotiate again and
again between different layers. That's a big waste and has very
low performance. We now only do the renegotiation for the highest
output layer. For other non output layers, we just keep a internal
buffer pool which is big enough to handle the surface allocation.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2382>
As SPEC says, when multi spatial layer exists, we should only output
one frame with the highest spatial id from each TU. We now store the
highest spatial layer information in the base class in order to let
the sub class handle different layers easily.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2382>
doesn't align on 20 millisecond frame size.
The AMR-WB codec imposes a fixed 20 millisecond frame size. In its current
form, the `voamrwbenc` plugin deals with this limitation by discarding any
audio at the end of the stream that falls short of 20 milliseconds. This patch
keeps the audio data, and appends silence to the end to preserve frame size
alignment.
The patch also adds tests to check for the updated behavior. I noticed that
tests weren't being built, so I changed the build to allow for building the
tests when the `tests` and `voamrwbenc` options are set.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3027>
when expose-all=False
When trying to find an decoder in that case, we loop over the different
decoder factories, and check that it outputs a format that matches the
requested one (through the :caps property), but if we find a decoder
that do match but later on some other don't we end up failing
autopluging. This patch ensures that we still plug the decoder that can
work.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3011>
- Update the docker image we use, starting using the standard one adding
`gtk4-doc` as required by rust plugins
- Update the plugins_doc_caches as required, some more plugins are built
with the new image
- Install ninja from pip as the version from F31 is too old
- Avoid buildings all GSreamer plugins when building the doc as it takes
time and resources for no good reason
- Stop linking to `GInstanceInitFunc` as it is not present in latest GLib
documentation, leading to warnings in hotdoc.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2954>
We are supposed to guarantee that pads that are exposed have the caps
set, but for sources that have pad with "all raw caps" templates, we end
up exposing pads that don't have caps set yet, which can break code (in
GES for example).
To avoid that we let uridecodebin plug a `decodebin` after such pads and
let decodebin to handle that for us. In the end the only thing that
decodebin does in those cases is to wait for pads to be ready and expose
them, after that `uridecodebin` will expose those pads.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3009>
We are always building our printf implementation, even when
GST_DEBUG is disabled, since we are exposing api (gst_print*)
that's dependant on our printf behavior.
We don't need to keep __gst_info_fallback_vasprintf around anymore.
Close#640
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/739>
Handle d3d11 device context in set_context() method with
additional device compatibility check so that only NVIDIA GPU
associated d3d11 device can be configured in the element.
And clear old d3d11 device per set_info() for d3d11 device to be
updated as well.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3018>
... and fix d3d11 specific enum type name
GST_CUDA_HAS_D3D is a build time define which indicates whether
GstD3D11 library is available or not, but DirectX SDK headers
must be available on the build system already.
Expose Direct3D related symbols if the build target is Windows
(i.e., if G_OS_WIN32 is defined)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3018>
GLib made the unfortunate decision to prevent libgobject from ever being
unloaded, which means that now any library which registers a static type
can't ever be unloaded either (and any library that depends on those,
ad nauseam).
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/778>
GstVA is not currently build by CI, because libva version is lower
than expected. So, the gstva library is not build, thus some symbols
aren't documented, breaking the documentation CI.
To move things forward, let's just remove temporarly the va plugins
from cache. While we decide on how to update the libva package in
the CI.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1025>
When advancing fragment in live, it's normal to return
GST_FLOW_EOS when playing at the live edge of the available
fragments. In that case, we still want to adjust bitrate
dynamically.
Fixes issue with dashdemux2 where the current bitrate of
each adaptation set is changed to the lowest one when
updating the mpd for a live stream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3020>
Instead of trying to hardcode site-packages paths for different platforms
just use python.get_install_dir() from meson and let it deal with the rest.
Also no longer try to import pygobject, which would otherwise not be
required at build time.
python.get_install_dir() was at the beginning broken on Windows, but
that was fixed in 0.60 via https://github.com/mesonbuild/meson/pull/9156
and since ges now requires >0.60 this can be ignored.
This change was motivated by the install path being wrong under MSYS2, where
the unix install layout is used and the detection code not taking that into
account.
This MR is a continuation of https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/230
see the discussion there for extra context.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3012>
Just like for the seconds field, there are no limitations on the hours and
minutes fields. The specification for xml schema duration fields doesn't forbid
specifying durations with only (huge) minutes or hours values.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2951>
When picking an available payload type, we need to pick one that is
available across all media.
The previous code, when multiple media were present, looked at the first one,
noticed it had pt 96 as the media pt, then simply looked at the next media,
noticed it didn't, and decided 96 was available.
Instead, check if the pt is used by any of the media, if it is, decide
it is not available and go to the next pt. I'm fairly sure that was the
original intent.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2984>
When updating a manifest during live playback, preserve the current
representation for each stream.
During update_fragment_info, if the current representation changed
because it couldn't be matched, trigger a caps change and new
header download.
This reverts commit e0e1db212f
and reapplies "dashdemux: Fix issue when manifest update sets slow start
without passing necessary header & caps changes downstream" with
changes.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2920>
Timers for RTX packets are dealt with later in update_rtx_timers(), and
timers for non-RTX packets would potentially also be unscheduled a
second time from there so avoid that.
Also don't shadow the timer variable from the outer scope but instead
make use of it directly.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2973>
Check back pressure of a stream transport before popping buffer from its backlog.
If the stream transport is not experiencing back pressure, the buffer can be popped from backlog and pushed to client.
Fixes:#1298
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2936>
The address/port is pre-defined by the caller of the function, so
retrying is only going to loop forever.
Ideally the multicast address should be checked after allocating but
this doesn't happen currently, so it's better to error out cleanly then
to loop forever trying the same address.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2975>
This avoids getting in a bunch of corner cases. We'd have to insert
a "rejected" line from the start as a place-holder to get around this,
but the rest of the code just becomes more complicated, so just
disallow it for now.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2439>
If the buffer is not msdk_buffer, we can try to directly import the
attached memory (i.e. va mem and dmabuf mem) by applying the common
uitl function: import_to_msdk_function ().
Here add a flag "from_qdata" in GstMsdkSurface to handle the cropping case,
we should avoid updating the crop values when msdk_surface is from the
memory's qdata, because the crop info from this surface is the already
updated one.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2498>
When input buffer is of dmabuf memory but not a msdk buffer (i.e., the
allocator is not msdk_allocator), then we can try to get fd of this mem,
create the corresponding va surface and wrap it as mfx surface.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2498>
If read_one or write_one was called but the stream closed before it could
read/write a whole packet, read_one/write_one would hang indefinitely,
consuming 100% CPU. This commit fixes that by treating a short read/write
as an error.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2964>
We were checking possible bind flags for the DXGI format
of the source texture but that's never applied to
the destination texture desc.
Just use the already configured bind (and misc) flags of source texture
for the destination texture allocation without additional check.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2950>
Setting the content-type property shall override internally detected MIME
types, to make it possible to do as following example (where audio/basic to be
used prior to audio/x-mulaw):
gst-launch-1.0 ... ! mulawenc ! audio/x-mulaw,rate=8000,channels=1 !
curlhttpsink location=<url> content-type=audio/basic
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2732>
* Private header name is changed to gstd3d11-private.h to follow
naming convention
* Add Since mark everywhere
* Update member variable names to be consistent with the other
object implementations in this library
* Correct outdated documentation
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2945>
Theoretically having elements in locked state should not have any effect
at all when the surrounding bin is doing state changes. However
previously a state change error of a locked element would cause the
bin's state change to also fail, which is clearly not intended.
State change failures of locked elements are to be handled by whoever
set the element to locked state. By always returning them here it is
impossible for the owner of the element to handle state change failures
gracefully without potentially affecting the whole pipeline's state
changes.
Non-failure returns are still returned as-is as the distinction between
ASYNC/NO_PREROLL/SUCCESS has big consequences on the state changes of
the bin and overall pipeline. Theoretically SUCCESS should also be
returned in all cases but I can't estimate the effects this would have
on the overall pipeline.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2932>
Current default G_MAXINT is not a correct value under any circumstances.
This creates an issue with screen capture, during which we currently do
not get any framerate info causing G_MAXINT to show up, where elements
downstream can possibly misbehave - for example, `vtenc` causes
a kernel panic.
Replace with 30/1 to avoid such scenarios.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2944>
The current handle_frame() does not return the real error that happens
in decode_scan and decode_frame, which makes the pipeline continue with
the error and may trigger asserting later.
We also return the error when decode_quant_table or decode_huffman_table
fails.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2938>
Add an example to show the usage of present singal.
In this example, a text overlay with alpha blended background
will be rendered on swapchain's backbuffer by using
Direct3D11, Direct2D, and DirectWrite APIs.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2923>
The "present" signal will be emitted just before the
IDXGISwapChain::Present() call. The client can perform additional
GPU operation with given GstD3D11Device object and
ID3D11RenderTargetView handle. Or, the client can read back
the scene to be displayed on window using the signal.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2923>
libsoup 3.0.x dispatches using a single source attached when the session
is created, so we need to create the session with the same context that
our download thread is later using.
2.74 or 3.1 will dispatch a response using the context which sent the
request. However, for any context other than the one that created the
session, this will also create and destroy sources, so there's still
some slight performance benefit.
Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1384
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2913>
This allows an application to provide their own opened DRM device
fd handle to kmssink. For example, an application can lease
multiple fd's from a DRM master to display on different CRTC
outputs at the same time with multiple kmssink instances.
Specifying the fd property is not allowed when driver-name
and/or bus-id properties are specified.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2807>
Handle select-streams and seek events in an element
level send_event() vfunc, so they can be received
before any source pads are created.
This allows preferred streams to be selected before
segment downloading starts.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2912>
Nouveau driver currently only exposes postproc entry. But
vaapidecodebin is registered independent if there are decoders or not,
exposing a segmentation fault.
This patch removes the encoder/decoder/codec arrays if no entries are
found, and if no decoders are found vaapidecodebin is not
registered. Also for vaapipostproc if no postproc entry is found.
Also, if general decoder, used by vaapidecodebin, doesn't have a sink
pad string, don't register the glib type.
Fixes: #1349
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2865>
Without this change cleanup function for g_autoptr is not defined for
GstPlayMediaInfo, GstPlaySignalAdapter, GstPlayVideoRenderer,
GstPlayVideoOverlayVideoRenderer and GstPlayVisualization. Cleanup
function was defined in gstplay.h, but missing in other header files.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2888>
When stopping the element, make sure the pad task
is stopped before destroying the part readers.
Closes a race where the pad task might access
a freed pointer.
Also add a guard against this sort of thing
by holding a ref to the reader in the pad loop.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2901>
When playing live, it's possible that one stream reaches
the end of the available playback window and goes to sleep
waiting for a manifest update, and the manifest update
introduces a new period. In that case, the sleeping
stream needs to wake up and go 'properly' EOS before we
can advance the input to the new period.
Accordingly, make sure that a stream's last_ret value
is not marked as EOS if it's just sleeping waiting for a live
manifest update.
Also fix the output loop to go back and re-check if it's
time to switch to the next period after dequeuing and
discarding an EOS event.
https://livesim.dashif.org/livesim/periods_20/testpic_2s/Manifest.mpd
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2895>
The parent refcount is of the *transformed* buffer, not the input
buffer.
Also update the docs to clarify that @transbuf is the transformed
buffer, and not the buffer on which a transformation is being
performed.
Due to this bug, modifying the structure of a meta that has been
copied to another buffer fails with:
gst_structure_set: assertion 'IS_MUTABLE (structure) || field == NULL' failed
Add a test for the same.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2890>
Newer compilers ( clang 15 ) have turned stricter and errors out instead
of warning on implicit function declations
Fixes
gstssaparse.c:297:12: error: call to undeclared library function 'isspace' with type 'int (int)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
while (isspace(*t))
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2879>
Ideally new() functions should simply call g_object_new() and not much
else, so let's do that here and handle all the construction properly in
a GObject way.
Now a play object created via g_object_new() is actually usable.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2880>
Ideally new() functions should simply call g_object_new() and not much
else, so let's do that here and handle all the construction properly in
a GObject way.
Now a player object created via g_object_new() is actually usable.
In addition, also fix the video-renderer property so that reading it
returns an object of the correct type.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2880>
That is, get rid of unnecessary and wrong special-casing.
This could always use gst_rtsp_url_get_request_uri_with_control() but as
we only have the control base URI as string it is easier to just call
gst_uri_join_strings().
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2868>
This was showing up as a memory leak in GTK's
gstreamer media backend:
40 bytes in 1 blocks are definitely lost in loss record 18,487 of 40,868
at 0x484586F: malloc (vg_replace_malloc.c:381)
by 0x50D5278: g_malloc (gmem.c:125)
by 0x50EDBA5: g_slice_alloc (gslice.c:1072)
by 0x50EFBCC: g_slice_alloc0 (gslice.c:1098)
by 0x51F2F45: g_type_create_instance (gtype.c:1911)
by 0x51DAE37: g_object_new_internal (gobject.c:2011)
by 0x51DC080: g_object_new_with_properties (gobject.c:2181)
by 0x51DCB20: g_object_new (gobject.c:1821)
by 0x9855F86: UnknownInlinedFun (gstplayer-wrapped-video-renderer.c:109)
by 0x9855F86: gst_player_new (gstplayer.c:579)
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1374
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2875>
Otherwise we won't send the protection packets for the last few
packets when a stream ends.
Also send EOS on the FEC src row pad immediately, and on the FEC src
column pad after draining is complete. This makes it so that the FEC
src pads on rtpbin behave the same way as the RTCP src pads on rtpbin
when EOS is received on the send_rtp_sink pad.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2863>
When a new segment event arrives, it immediately updates
the current stored segment, which was used for calculating
the running time of the current text buffer for every
passing video frame. This means a segment that arrives
after the text buffer might get used to (mis)calculate
the running times subsequently.
Instead, calculate and store the right running time
using the current segment when storing the buffer. Later
the stored segment can get freely updated.
This fixes the case where pieces of video and text streams
are seamlessly concatenated and fed through the text overlay.
Previously, it could lead to the current text buffer suddenly
have a massive running time and blocking all further input.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2802>
Radeon mesa gallium driver has a bug which adds P010_10LE sink caps
format. This patch removes formats which arent 420 chroma.
gst_caps_set_format_array() wasn't used because the fix traverse
several structures with potential different formats.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2844>
When returning GST_ADAPTIVE_DEMUX_FLOW_RESTART_FRAGMENT
for the first segment data, we might need to requeue the
header.
This was leading to occasional prerolling stalls on
HLS live streams with renditions.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2849>
Make sure gst_adaptive_demux_loop_cancel_call()
never tries to operate on an invalidated main context. Make
sure to clear the main context pointer while holding the lock,
and to check it in gst_adaptive_demux_loop_cancel_call()
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2847>
GLib's GRecMutex will allocate another heap memory for CRITICAL_SECTION
struct and g_rec_mutex_lock/g_rec_mutex_unlock use WIN32 APIs actually.
We don't need such intermediate function calls and redundant heap allocation.
Just call WIN32 APIs directly.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2845>
Media playlist updates and fragment downloads happen in an interleaved
fashion. When a media playlist update fails *while* a segment is being
downloaded, this means we lost synchronization.
Properly propagate and handle this
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2839>
There is now only a single case where we setup the initial playlist to 0, which
is for the very first variant stream.
Rendition streams will have the initial playlist "synchronized" against the
variant stream media playlist.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2839>
Loss of synchronization happens when the updated media playlist has no
relationship to the previous ones. This could happen because of network issues,
server issues, etc...
When this happens, we take no chance and "reset" ourselves so that we can "seek
back to live" against the new updated playlists.
Since this happens at the "media playlist update" level, make sure the custom
flow return is propagated up.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2839>
We are already in the main scheduler thread, therefore we can do the "seek back
to live" directly. This also avoids other pending actions to take place.
Also handle the loss of sync when doing manifest updates.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2839>
Close some race conditions in switching to the next period,
by ensuring the tracks are completely drained first and by
not outputting EOS events to the output source pad
if there is another period pending.
Fixes Manifest_MultiPeriod_1080p.mpd some more.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2838>
Before sending EOS, update the period's has_next_period
flag and/or create the next period. This closes a race
where the output loop might receive the EOS event
and either push it downstream (causing premature EOS),
or receive it and try and switch to the next period
before that period is completely set up.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2838>
When combining stream flows, ignore streams that
are not selected, instead of checking whether
the stream state has changed yet.
Fixes another issue with dashdemux2 where it fails to
change to the next period when playing content with
several video, audio and text streams, as with
Manifest_MultiPeriod_1080p.mpd when seeking to 730
just before the end of the first period.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2838>
The is_gst_mini_object_check would sometimes detect a proper GObject
as a mini object, and then bad things happen.
We know whether a pointer is a proper GObject or a MiniObject here
though, so just pass that information to the right code paths and
avoid the heuristics altogether.
Eliminates all remaining uses of object_is_gst_mini_object().
Fixes#1334
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2832>
The is_gst_mini_object_check would sometimes detect a proper GObject
as a mini object, and then bad things happen.
We know whether a pointer is a proper GObject or a MiniObject here
though, so just pass that information to the right code paths and
avoid the heuristics altogether.
There are probably more cases where the check should be eliminated.
Fixes#1334, maybe
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2832>
