This doesn't always bring visible issue, but is formally incorrect:
not chaining up means that the code doesn't trigger GstObject and
GstElement "constructed" implementations.
In particular both GstElement's and GstObject's classes in
"constructed" may sign up this object for tracing and
GstObject's class sets GST_OBJECT_FLAG_CONSTRUCTED flag.
If we don't chain up none of this is going to be executed.
For example, before the fix leaks tracer couldn't detect this leak:
```c
int main (int argc, char **argv) {
g_setenv ("GST_TRACERS", "leaks(name=all-leaks)", TRUE);
g_setenv ("GST_DEBUG", "GST_TRACER:7", TRUE);
g_setenv ("G_DEBUG", "fatal-warnings", TRUE);
gst_init (&argc, &argv);
// leak audiomixer: doesn't detect because it's based on the aggregator
gst_element_factory_make ("audiomixer", "Jerry");
// leak videoconvert: this one is detected fine because it's not
// based on the aggregator
//gst_element_factory_make ("videoconvert", "Tom");
gst_deinit ();
return 0;
}
// $ cc tst.c $(pkg-config --cflags --libs gstreamer-1.0) -o tst && ./tst
```
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8416>
When the stream resolution change it is needed to negotiate
a new pools and to update the caps.
Resolution change could occurs on a new sequence or a new
picture so move resolution change detection code in a common
function.
Only call streamoff if the resolution occur while decoding a key frame.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8064>
We must drain the pending output picture so that subclass can renegotiate
the caps. Not doing so while still renegotiating would mean that the
subclass would have to do an allocation query before pushing the caps.
Pushing the caps now without this would also not work since these caps
won't match the pending buffers format.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8064>
With the old method of adding build RPATHs, only the libsoup as
a subproject was supported. However, it's possible to use Homebrew for
libsoup too by adding libsoup as a dependency to the target (even
though we don't link to it).
Apple ld will not add a LC_LOAD_DYLIB entry for a library unless there
are symbols that need it because meson passes `-dead_strip_dylibs`, so
we can do this without issue.
As a bonus, this ensures the correct build RPATHs in all cases.
Additionally, we still need to explicitly add an LC_RPATH for the
installed case.
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4027
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8403>
In many places in the log tracer %d was used for data types we have
string format functions. This is especially problematic when the data
type is not immediately obvious to the reader (e.g. gboolean vs
GstFlowRet, where TRUE==1 but GST_FLOW_OK==0) or the values are not easy
to memorize (GST_STATE_CHANGE_PLAYING_TO_PAUSED==35).
This patch replaces all the integer codes with string representations so
that the logs are easier to understand by humans.
The formatting of GstBufferList by the log tracer is also updated to use
GST_PTR_FORMAT instead of a raw pointer, so the user can see the
timestamps, size and number of buffers.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8414>
For VPS, PPS, APS, OPI and DCI, the extension flags are the last syntax
in the structures, and according to the spec, should be ignored if set to 1.
Therefore, we can just ignore them rather than failing.
This fixes a few failures in fluster, like in the PSEXT_A_Nokia_2 stream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8359>
If we find VPS/SPS/PPS in codec_data and call gst_h266_parse_process_nal
with them, we need to have negotiated before in order to correctly
process them with flags like h266parse->transform set or not depending
on the negotiation. This is important because in certain vvc1/vvi1 streams we
may have correct codec_data but faulty parameter sets in the stream and
we would want to push the parameter sets from codec_data first.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8359>
implement serialization of codec_data containing VvcDecoderConfigurationRecord
as defined in ISO/IEC 14996-15.
The VPS/SPS/PPS NALs are added to the codec_data. APS NALs could be
optionally included as well but will be pushed in-band instead, because:
1. Logic is easier that way. We'd have to filter out for PREFIX_APS only
(SUFFIX_APS aren't allowed in codec_data).
2. APS NALs can also be sent for every non-keyframe slice, and often are, so just pushing
them in-band makes more sense to have less to keep track and avoid possible
duplicates.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8359>
If collecting the backchannel stream after the stream has begun
preparing, make sure to join it to the RTSP media bin, as is
done for dynamic payloaders in the pad-added handler. Add a
new internal stream creation method for it, that could be
publically exposed later if it proves useful.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8378>
When prerolling / blocking an RTSP media, don't consider receiver
streams like onvif backchannel in the 'is everything blocked' check,
as those streams won't receive data until after the RTSP stream
is running.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8378>
When decoupling send/receive loops in rtsp-media using appsink/appsrc,
put the appsink in the bin before linking it, as this code path
had that wrong.
For both sender and receiver paths, if the pipeline already started
also add the other appsrc/appsink of the pair to the pipeline
immediately instead of putting it in the pending elements list,
as they won't get added later.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8378>
The example was switched to use autovideosink, which broke
current position calculation based on the last video frame.
Add code to retrieve the actual video sink from within autovideosink
to make it work again.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8378>
Suppose you're trying to debug why this pipeline doesn't work:
```
$ GST_DEBUG=GST_PIPELINE:DEBUG gst-launch-1.0 \
videotestsrc num-buffers=10 ! x264enc name=enc ! mux.sink_0 \
mpegtsmux name=mux ! fakesink
```
You will encounter this line in the logs:
> gst_parse_perform_link: linking some pad of GstX264Enc named enc to
> pad mux of GstMpegTsMux named mux (0/1) with caps "(NULL)"
It would seem that the element name is being read as a pad name as well,
and that made me wonder if the parsing was not working. However, it was
just a bug in the code printing that log. This patch fixes that bug.
Note that it is possible to specify more than one pad name for each side
of the link. For instance, the following is a valid pipeline that will
remux the video and audio of an MP4 file into MKV:
```
$ GST_DEBUG=GST_PIPELINE:DEBUG gst-launch \
filesrc location=input.mp4 ! qtdemux name=demux \
multiqueue name=mq \
matroskamux name=mux ! filesink location=output.mkv \
demux.video_0,audio_0 ! mq.sink_0,sink_1 \
mq.src_0,src_1 ! mux.video_0,audio_0
```
The new logging accomodates this by using a new utility function to join
strings of pad name lists instead of `PRETTY_PAD_NAME_FMT` (which only
supports one pad name). For example:
> linking pads {video_0, audio_0} of GstQTDemux named demux to pads
> {sink_0, sink_1} of GstMultiQueue named mq with caps "(NULL)"
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8415>
"undefined reference to `GST_TRACER_PAD_SEND_EVENT_PRE'
undefined reference to `GST_TRACER_PAD_SEND_EVENT_POST'"
errors are generated when trying to build GStreamer with
the following build configuration:
meson setup -Dgstreamer:tracer_hooks=false build
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8419>
Similar to qtdemux.
Tested against other Gst elements and MPV. Note that the later
apparently does not show correct results for flipped values.
In particular the Yaw value seems to get ignored by many clients.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8319>
Similar to qtmux, but for mkv and webm containers.
Tested against other Gst elements and MPV. Note that the later
apparently does not show correct results for flipped values.
In particular the Yaw value seems to get ignored by many clients.
Can be tested with:
```
gst-launch-1.0 \
videotestsrc num-buffers=90 ! \
taginject tags="image-orientation=rotate-270" ! \
capsfilter caps=video/x-raw,width=640,height=480,max-framerate=30/1 ! \
videoconvert ! \
queue ! \
vp8enc ! \
queue ! \
webmmux ! \
filesink location=./test.webm
```
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8319>
Sometimes the seek to the end of file to determine the duration would trigger a
reset of the source pads, that would confuse the decoder downstream and trigger
an error. So the proposed fix is to not reset pads when the segment event being
processed is the consequence of a seek performed to determine the duration.
Fixes#4212
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/937>
For outputs with a high number of channels, macOS has a bug where
initially CoreAudio will report incorrect positions for all channels,
but after you run Audio MIDI Setup and configure the speaker layout
there, macOS will always report those few as positioned, with no option
to revert that (other than deleting some internal files).
In such scenario our code would just ignore all the unpositioned
channels. Since you can only position max. 16 channels in macOS, if you
had more on your output device, those would be unusable.
This commit makes sure that in addition to the usual positioned layout
(if there is one), we will expose caps for a no-positions layout that
always has the maximum amount of channels available.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8311>
By default, for devices with larger amounts of outputs, CoreAudio can
provide invalid channel labels/positions, simply by starting at 0 and
incrementing forward. For example, values 19 through 32 are not valid
according to the CoreAudioBaseTypes.h header, but if your device has >19
output channels, you will find CoreAudio using those values.
This is most likely a bug in CoreAudio, since in that case it should use
unpositioned labels (e.g. _Discrete_X) instead.
This commit aims to work around this by overriding all channels to be
unpositioned if the case above is detected.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8311>
After a bit of back and forth, we figure-out that for backward
compatibility we need to set the tile stride the way GStreamer
defines it. Sinks such as glimagesink/waylandsink translate it
back to the number of bytes representation used by Linux.
The change in !7355 when the other way around, breaking tiled
playback through waylandsink and glimagesink.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7892>
Mirroring the demux element and isomp4mux from gst-plugins-rs.
Tested against other Gst elements and MPV. Note that the later
apparently does not show correct results for flipped values.
Can be tested with:
```
gst-launch-1.0 \
videotestsrc num-buffers=90 ! \
taginject tags="image-orientation=rotate-90" ! \
capsfilter caps=video/x-raw,width=640,height=480,max-framerate=30/1 ! \
videoconvert ! \
queue ! \
openh264enc ! \
queue ! \
h264parse ! \
mp4mux ! \
filesink location=./test.mp4
```
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8371>
There is no requirement for a base DRM format to be supported by libgstvideo
in order to be uploaded to.
The linux-dmabuf-v1 format events are DRM_FORMAT codes and don't need to
be converted before use with `gst_video_dma_drm_fourcc_to_string`.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8279>
