This commit adds an Android `webrtcsrc` based example with the following
features:
* A first view allows retrieving the producer list from the signaller (peer ids
are uuids which are too long to tap, especially using an onscreen keyboard).
* Selecting a producer opens a second view. The first available video stream is
rendered on a native Surface. All the audio streams are rendered using
`autoaudiosink`.
Available Settings:
* Signaller URI.
* A toggle to prefer hardware decoding for OPUS, otherwise the app defaults to
raising `opusdec`'s rank. Hardware decoding was moved aside since it was found
to crash the app on all tested devices (2 smartphones, 1 tv).
**Warning**: in order to ease testing, this demonstration application enables
unencrypted network communication. See `AndroidManifest.xml`.
The application uses the technologies currenlty proposed by Android Studio when
creating a new project:
* Kotlin as the default language, which is fully interoperable with Java and
uses the same SDK.
* gradle 8.6.
* kotlin dialect for gradle. The structure is mostly the same as the previously
preferred dialect, for which examples can be found online readily.
* However, JNI code generation still uses Makefiles (instead of CMake) due to
the need to call [`gstreamer-1.0.mk`] for `gstreamer_android` generation.
Note: on-going work on that front:
- https://gitlab.freedesktop.org/gstreamer/cerbero/-/merge_requests/1466
- https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6794
Current limitations:
* x86 support is currently discarded as `gstreamer_android` libs generation
fails (observed with `gstreamer-1.0-android-universal-1.24.3`).
* A selector could be added to let the user chose the video streams and
possibly decide whether to render all audio streams or just select one.
Nice to have:
* Support for the synchronization features of the `webrtc-precise-sync-recv`
example (NTP clock, RFC 7273).
* It could be nice to use Rust for the specific native code.
[`gstreamer-1.0.mk`]: https://gitlab.freedesktop.org/gstreamer/cerbero/-/blob/main/data/ndk-build/gstreamer-1.0.mk
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1578>
Only configure header extensions from the source pad caps if they exist
already in the source pad caps, otherwise the configuration will fail.
Extensions that are added via the signals might not exist in the source
pad caps yet and would be added later.
Also, if configuring an existing extension from the new caps fails,
remove it and try to request a new extension for it.
Additionally don't remove extensions from the caps that can't be
provided. No header extensions for them would be added to the packets,
but that's not a problem. Removing them on the other hand would cause
negotiation to fail. This only affects extensions that are already
included in the caps.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1577>
If configuring an existing extension from the new caps fails, remove it
and try to request a new extension for it.
Also remove all extensions from the list that are not provided in the
caps, instead of passing RTP packets to all of them anyway.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1577>
In our tests, the slope (found with linear regression) on a
history of the (smoothed) accumulated inter-group delays
gives a more stable congestion control. In particular,
low-end devices becomes less sensitive to spikes in
inter-group delay measurements.
This flavour of delay based bandwidth estimation with Google
Congestion Control is also what Chromium is using.
To make it easy to experiment with the new estimator, as
well as add support for new ones in the future, also add
infrastructure for making delay estimator flavour selectable
at runtime.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1566>
Rust targets without support for `AtomicU64` is still
somewhat common. Running
git grep -i 'max_atomic_width: Some(32)' | wc -l
in the Rust compiler repo currently counts to 34 targets.
Change the `RtpBasePay2::ssrc_collision` from `AtomicU64` to
`Mutex<Option<u32>>`. This way we keep support for these
targets.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1562>
While monitoring and debugging rtpgccbwe, it is very helpful
to get continuous values of what it considers the effective
bitrate. Right now such prints will stop coming once the
algorithm stabilizes. Print it in more places so it keeps
coming. Use the same format to make it simpler to extract
the values by parsing the logs.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1567>
When debugging rtpgccbwe it is helpful to know if it is
delay based or loss based band-width estimation that puts a
bound on the current target bitrate, so add logs for that.
To minimize the time we need to hold the state lock, perform
the logging after we have released the state lock.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1561>
This set of changes implements the below fixes:
- Allow certificates to be specified for client/quicsink
- Secure connection being true on server/quicsrc and false on
client/quicsink still resulted in a successful connection
instead of server rejecting the connection
- Using secure connection with ALPN was not working
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1036>
If a user constrained the supported CAPS, for instance using `video-caps`:
```shell
gst-launch-1.0 videotestsrc ! video/x-raw,format=I420 ! x264enc \
! webrtcsink video-caps=video/x-vp8
```
... a panic would occur which was internally caught without the user being
informed except for the following message which was written to stderr:
> thread 'tokio-runtime-worker' panicked at net/webrtc/src/webrtcsink/imp.rs:3533:22:
> expected audio or video raw caps: video/x-h264, [...] <br>
> note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
The pipeline kept running.
This commit converts the panic into an `Error` which bubbles up as an element
`StreamError::CodecNotFound` which can be handled by the application.
With the above `gst-launch`, this terminates the pipeline with:
> [...] ERROR webrtcsink net/webrtc/src/webrtcsink/imp.rs:3771:gstrswebrtc::
> webrtcsink:👿:BaseWebRTCSink::start_stream_discovery_if_needed::{{closure}}:<webrtcsink0>
> Error running discovery: Unsupported caps: video/x-h264, [...] <br>
> ERROR: from element /GstPipeline:pipeline0/GstWebRTCSink:webrtcsink0:
> There is no codec present that can handle the stream's type. <br>
> Additional debug info: <br>
> net/webrtc/src/webrtcsink/imp.rs(3772): gstrswebrtc::webrtcsink:👿:BaseWebRTCSink::
> start_stream_discovery_if_needed::{{closure}} (): /GstPipeline:pipeline0/GstWebRTCSink:webrtcsink0:
> Failed to look up output caps: Unsupported caps: video/x-h264, [...] <br>
> Execution ended after 0:00:00.055716661 <br>
> Setting pipeline to NULL ...
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1540>
Clippy caught the missing feature `signal` which is used by the WebRTC precise
synchronization examples. When running `cargo` `check`, `build` or `clippy`
without `no-default-dependencies`, this feature was already present due to
dependents crates.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1541>
When swapping between several development branches, compilation times can be
frustrating. This commit proposes adding features to control which signaller
to include when building the webrtc plugin. By default, all signallers are
included, just like before.
Compiling the `webrtc-precise-sync` examples with `--no-default-features`
reduces compilation to 267 crates instead of 429 when all signallers are
compiled in.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1539>
This commit implements [RFC 7273] (NTP & PTP clock signalling & synchronization)
for `webrtcsink` by adding the "ts-refclk" & "mediaclk" SDP media attributes to
identify the clock. These attributes are handled by `rtpjitterbuffer` on the
consumer side. They MUST be part of the SDP offer.
When used with an NTP or PTP clock, "mediaclk" indicates the RTP offset at the
clock's origin. Because the payloaders are not instantiated when the offer is
sent to the consumer, the RTP offset is set to 0 and the payloader
`timstamp-offset`s are set accordingly when they are created.
The `webrtc-precise-sync` examples were updated to be able to start with an NTP
(default), a PTP or the system clock (on the receiver only). The rtp jitter
buffer will synchronize with the clock signalled in the SDP offer provided the
sender is started with `--do-clock-signalling` & the receiver with
`--expect-clock-signalling`.
[RFC 7273]: https://datatracker.ietf.org/doc/html/rfc7273
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1500>
When an encoder was not supported by the `VideoEncoder` `bitrate` accessors, an
`unimplemented` panic would occur which would poison `state` & `settings`
`Mutex`s resulting in other threads panicking, notably entering `end_session()`,
which lead to many failures in `BinImplExt::parent_remove_element()` until a
segmentation fault ended the process. This was observed using `vaapivp9enc`.
This commit logs a warning if an encoder isn't supported by the `bitrate`
accessors and silently by-passes `bitrate`-related operations when unsupported.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1534>
Otherwise the clock id will simply be overridden instead of unscheduling
it, and if the streaming thread of the source pad currently waits on it
then it will wait potentially for a very long time and deactivating the
pad would wait for that to happen.
Also unschedule the clock id on `Drop` of the state to be one the safe
side and not simply forget about it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1526>
Add `self.increasing_duration` and `self.increasing_counter`
to logs to provide more details of why `overuse_filter()`
determines overuse of network.
To get access to the latest values of those fields we need
to move down the log call. But that is fine, since no other
logged data is modified between the old and new location of
`gst::log!()`.
We do not bother logging `self.last_overuse_estimate` since
that is simply the previously logged value of `estimate`. We
must put the log call before we write the latest value to it
though, in case we want to log it in the future.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1522>
Quoting [`BehaviorVersion` documentation]:
> Over time, new best-practice behaviors are introduced. However, these
> behaviors might not be backwards compatible. For example, a change which
> introduces new default timeouts or a new retry-mode for all operations might
> be the ideal behavior but could break existing applications.
This commit uses `BehaviorVersion::v2023_11_09()`, which is the latest
major version at the moment. When a new major version is released, the method
will be deprecated, which will warn us of the new version and let us decide
when to upgrade, after any changes if required. This is safer that using
`latest()` which would silently use a different major version, possibly
breaking existing code.
[`BehaviorVersion` documentation]: https://docs.rs/aws-config/1.1.8/aws_config/struct.BehaviorVersion.html
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1520>
The following error is logged when `webrtcsink` is feeded with an audio stream:
> ERROR video-info video-info.c:540:gst_video_info_from_caps:
> wrong name 'audio/x-raw', expected video/ or image/
This commit bypasses `VideoInfo::from_caps` for audio streams.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1511>
Some elements in the RTP stack assume all buffers in a `gst::BufferList`
correspond to the same timestamp. See in [`rtpsession`] for instance.
This also had the effect that `rtpsession` did not create correct RTCP as it
only saw some of the SSRCs in the stream.
`rtpgccbwe` formed a packet group by gathering buffers in a `gst::BufferList`,
regardless of whether they corresponded to the same timestamp, which broke
synchronization under certain circonstances.
This commit makes `rtpgccbwe` push the buffers as they were received: one by one.
[`rtpsession`]: bc858976db/subprojects/gst-plugins-good/gst/rtpmanager/gstrtpsession.c (L2462)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1502>
Instead of exposing all ids properties as strings, we now have two
signaller implementations exposing those properties using their actual
type. This API is more natural and save the element and application
conversions when using numerical ids (Janus's default).
I also removed the 'joined-id' property as it's actually the same id as
'feed-id'. I think it would be better to have a 'janus-state' property or
something like that for applications willing to know when the room has
been joined.
This id is also no longer generated by the element by default, as Janus
will take care of generating one if not provided.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1486>
Without sending a Leave request to the server before disconnecting, the
disconnected client will appear present and stuck in the room for a little
while until the server removes it due to inactivity.
After this change, the disconnecting client will immediately leave the room.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1482>
It may be necessary for some signalling clients but the source element
doesn't need to depend on it.
Also, the value will fall back to the pad's MSID for the first argument
to the request-encoded-filter gobject signal when it isn't available
from the signalling client.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1477>
Don't use connect(), since that is incompatible with multicast.
Instead, drop received packets that are from senders we do not want.
Also set multicast loopback = false so we don't receive RTCP RRs from
ourselves and interpret them as RTCP SRs.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1425>
If we only send a single Transport in the Transports header, then the
server is allowed to omit it in the response. This has some strange
consequences for UDP transport: specifically, we have no idea what
addr/port we will get the packets from.
In those cases, we connect() on the socket when we receive the first
packet, so we can send RTCP RRs, and also so we can ensure that we
ignore data from other addresses.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1425>
GST_PLUGIN_FEATURE_RANK=rtspsrc2:1 gst-play-1.0 [URI]
Features:
* Live streaming N audio and N video
- With RTCP-based A/V sync
* Lower transports: TCP, UDP, UDP-Multicast
* RTP, RTCP SR, RTCP RR
* OPTIONS DESCRIBE SETUP PLAY TEARDOWN
* Custom UDP socket management, does not use udpsrc/udpsink
* Supports both rtpbin and the rtpbin2 rust rewrite
- Set USE_RTPBIN2=1 to use rtpbin2 (needs other MRs)
* Properties:
- protocols selection and priority (NEW!)
- location supports rtsp[ut]://
- port-start instead of port-range
Co-Authored-by: Tim-Philipp Müller <tim@centricular.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1425>
This change addresses a cosmetic issue with livekit, where the
connection quality indicator seen by other users shows bad quality
unless the track is added with a high quality layer. The details of the
layer submitted aren't significant for this purpose.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1443>
In the signaller clients and servers, the following sequence is used to close
the websocket (in the [send task]):
```rust
ws_sink.send(WsMessage::Close(None)).await?;
ws_sink.close().await?;
```
tungstenite's [`WebSocket::close()` doc] states:
> Calling this function is the same as calling `write(Message::Close(..))``
So we might think they are redundant and either could be used for this purpose
(`send()` calls `write()`, then `flush()`).
The result is actually is bit different as `write()` starts by checking the
state of the connection and [returns `SendAfterClosing`] if the socket is no
longer active, which is the case when a closing request has been received from
the peer via a [call to `do_close()`]). Note that `do_close()` also enqueues a
`Close` frame.
This behaviour is visible from the server's logs:
```
1. tungstenite::protocol: Received close frame: None
2. tungstenite::protocol: Replying to close with Frame { header: FrameHeader { .., opcode: Control(Close), .. }, payload: [] }
3. gst_plugin_webrtc_signalling::server: Received message Ok(Close(None))
4. gst_plugin_webrtc_signalling::server: connection closed: None this_id=cb13892f-b4d5-4d59-95e2-b3873a7bd319
5. remove_peer{peer_id="cb13892f-b4d5-4d59-95e2-b3873a7bd319"}: gst_plugin_webrtc_signalling::server: close time.busy=285µs time.idle=55.5µs
6. async_tungstenite: websocket start_send error: WebSocket protocol error: Sending after closing is not allowed
```
1: The server's websocket receives the peer's `Close(None)`.
2: `do_close()` enqueues a `Close` frame.
3: The incoming `Close(None)` is handled by the server.
4 & 5: perform session closing.
6: `ws_sink.send(WsMessage::Close(None))` attempts to `write()` while the ws
is no longer active. The error causes an early return, which means that
the enqueued `Close` frame is not flushed.
Depending on the peer's shutdown sequence, this can result in the following
error, which can bubble up as a `Message` on the application's bus:
```
ERROR: from element /GstPipeline:pipeline0/GstWebRTCSrc:webrtcsrc0: GStreamer encountered a general stream error.
Additional debug info:
net/webrtc/src/webrtcsrc/imp.rs(625): gstrswebrtc::webrtcsrc:👿:BaseWebRTCSrc::connect_signaller::{{closure}}::{{closure}} (): /GstPipeline:pipeline0/GstWebRTCSrc:webrtcsrc0:
Signalling error: Error receiving: WebSocket protocol error: Connection reset without closing handshake
```
On the other hand, [`close()` ensures the ws is active] before attempting to
write a `Close` frame. If it's not, it only flushes the stream.
Thus, when we want to be able to close the websocket and/or to honor the closing
handshake in response to the peer `Close` message, the `ws_sink.close()`
variant is preferable.
This can be verified in the resulting server's logs:
```
tungstenite::protocol: Received close frame: None
tungstenite::protocol: Replying to close with Frame { header: FrameHeader { is_final: true, rsv1: false, rsv2: false, rsv3: false, opcode: Control(Close), mask: None}, payload: [] }
gst_plugin_webrtc_signalling::server: Received message Ok(Close(None))
gst_plugin_webrtc_signalling::server: connection closed: None this_id=192ed7ff-3b9d-45c5-be66-872cbe67d190
remove_peer{peer_id="192ed7ff-3b9d-45c5-be66-872cbe67d190"}: gst_plugin_webrtc_signalling::server: close time.busy=22.7µs time.idle=37.4µs
tungstenite::protocol: Sending pong/close
```
We now get the notification `Sending pong/close` (the closing handshake) instead
of `websocket start_send error` from step 6 with previous variant.
The `Connection reset without closing handshake` was not observed after this
change.
[send task]: 63b568f4a0/net/webrtc/signalling/src/server/mod.rs (L165)
[`WebSocket::close()` doc]: https://docs.rs/tungstenite/0.21.0/tungstenite/protocol/struct.WebSocket.html#method.close
[returns `SendAfterClosing`]: 85463b264e/src/protocol/mod.rs (L437)
[call to `do_close()`]: 85463b264e/src/protocol/mod.rs (L601)
[`close()` ensures the ws is active]: 85463b264e/src/protocol/mod.rs (L531)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1435>
We were setting audio and video caps by default even when the user
might have requested only video or audio. This would then result
in a `Could not reuse transceiver` error from the webrtcbin.
Fix this by allowing the user to specify audio or video caps as
None. This allows us to maintain the earlier behaviour for backward
compatibility while allowing the user to not request audio or video
as need be.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1433>
When streaming small amounts of data, using awss3sink might not be a
good idea, as we need to accumulate at least 5 MB of data for a
multipart upload (or we flush on EOS).
The alternative, while inefficient, is to do a complete PutObject of
_all_ the data periodically so as to not lose data in case of a pipeline
failure. This element makes a start on this idea by doing a PutObject
for every buffer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1337>
webrtcbin will refuse pad requests for all sorts of reasons, and should
be logging an error when doing so, simply post an error message and let
the application deal with it, the reason for the refusal should
hopefully be available in the logs to the user.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1399>
Implement new signaller WhipServerSignaller
- an http server using 'warp'
- handlers for the POST, OPTIONS, PATCH and DELETE
- fixed path `/whip/endpoint` as the URI
- fixed value 'whip-client' as the producer peer id
- fixed resource url `/whip/resource/whip-client`
Derive whipserversrc element from BaseWebRTCSrc
- implement constructed method for ObjectImpl to set
non-default signaller, i.e., WhipServerSignaller
- bind the properties stun-server and turn-servers to those on
the Signaller
Connect to 'webrtcbin-ready' signal in the constructor of WhipServerSignaller
- it will be emitted by the webrtcsrc when the webrtcbin element is ready
- the closure for this signal will in turn connect to webrtcbin's ice-gathering-state
and perform send with the answer sdp via the channel
- the WhipServer will hold its HTTP response in POST handler until this signal
is received or timeout which happens early
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1284>
add a new signal webrtcbin-ready in this place doing same
thing but can be used for both consumers and producers
Please note this change is only to the consumer-added
signal on the signaller interface.
The consumer-added signal on the webrtcsink is unchanged
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1284>
Also move processing from the capture thread to the streaming thread.
The NDI SDK can cause frame drops if not reading fast enough from it.
All frame processing is now handled inside the ndisrcdemux.
Also use a buffer pool for video if copying is necessary.
Additionally, make sure to use different stream ids in the stream-start
event for the audio and video pad.
This plugin now requires GStreamer 1.16 or newer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1365>
If a packet is starting with a leading fragment but we do not expect to
receive one, then skip over it to the next OBU.
Not doing so would cause parsing of the middle of an OBU, which would
most likely fail and cause unnecessary warning messages about a
corrupted stream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1367>
The "signaller" property used to be defined as MUTABLE_READY which meant that
the property was always set after `constructed()` was called.
Since `connect_signaller()` was called from `constructed()`, only the default
signaller was used.
This commit sets the "signaller" property as CONSTRUCT_ONLY. Using a builder,
this property will now be set before the call to `constructed()`.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1324>
During `on_remote_description_set()` processing, current session is removed
from the sessions `HashMap`. If an ice candidate is submitted to `handle_ice()`
by that time, the session can't be found and the candidate is ignored.
This commit wraps the Session in the sessions `HashMap` so an entry is kept
while `on_remote_description_set()` is running. Incoming candidates received by
`handle_ice()` will be processed immediately or enqueued and handled when the
session is restored by `on_remote_description_set()`.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1325>
The NDI closed captions specifications [1] define a variation where metadata is
attached to the video frame. This requires the AFD buffer to be v210 encoded.
This commit applies this strategy.
Another difference with previous version is that when an error occurs while
encoding or decoding a meta, next meta are also tried instead of failing
immediately.
Receiving closed captions as a standalone metadata is kept for interoperability
purposes. In this case, metadata is also expected to be v210 encoded.
[1]: http://www.sienna-tv.com/ndi/ndiclosedcaptions.html
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1356>
@to-owned increases refcount of the element, which prevents the object from proper destruction, as the initial refcount with ElementFactory::make is larger than 1.
Instead, use @watch to create a weak reference and unbind the closure automatically if the object gets destroyed
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1355>
