It will likely confuse players. We can drop corresponding
cue since the text line will be included in the next cue

Sebastian Dröge authored Sep 28, 2022 and Sebastian Dröge committed Sep 28, 2022

rtpav1: Use GStreamer types by namespace instead of importing dozens of types directly into the scope

For consistency with other plugins and to reduce confusion of where
types actually come from.

... as described in the spec

Sebastian Dröge authored Sep 27, 2022 and Sebastian Dröge committed Sep 27, 2022

Otherwise EOS might be sent before the last frame's data, or even at a
much earlier frame due to reordering.

Otherwise this can potentially deadlock or delay state changes for a
longer time.

The order is first stream lock, then state lock. Everything else can
lead to deadlocks.

So the online documentation works

Always first try draining queued data in the loop and only start waiting
if there's nothing to drain right now. Otherwise data might have to be
drained right now but we still wait and nothing is ever waking up the
source pad task again.

Also make sure to not wait multiple times on the same gst::ClockId but
instead unset it after waiting on it and no new one was scheduled in the
meantime. Future waits on the same ClockId will immediately return and
instead we should wait on the condvar if no new ClockId is available.

Otherwise receiving StreamStart after Eos might keep the source pad task
paused and no new data is ever pushed downstream.

Now we don't need to keep anything in sync manually.

Also fix the images_template.yml include while we're at it.

Part-of: <!875>

Part-of: <!874>

Part-of: <!874>

It's in the same repository after all.

Mart Raudsepp authored Sep 19, 2022 and Arun Raghavan committed Sep 19, 2022

!698
raised the dependency via bumping the dav1d rust crate used, but didn't add
a requirement at meson level, thus with automatic or enabled option for dav1d
it would pass with an older failure, but then during compilation phase fail
with:

  --- stderr
  thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: PkgConfig(`"pkg-config" "--libs" "--cflags" "dav1d" "dav1d >= 1.0.0"` did not exit successfully: exit status: 1
  error: could not find system library 'dav1d' required by the 'dav1d-sys' crate

  --- stderr
  Package dependency requirement 'dav1d >= 1.0.0' could not be satisfied.
  Package 'dav1d' has version '0.8.2', required version is '>= 1.0.0'
  )', /home/user/.cargo/registry/src/github.com-1ecc6299db9ec823/dav1d-sys-0.5.0/build.rs:80:10
  note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
warning: build failed, waiting for other jobs to finish...

Thibault Saunier authored Sep 17, 2022 and Sebastian Dröge committed Sep 19, 2022

We were globing recursively during meson run and it was spending 20secs
here in total only to run the dependencies.py script

hlssink can be built by default because it has no dependencies.

tutorial and rsfile should not be built by default because they are
not very useful, and flavors should not be built by default because
it's very incomplete.

They would potentially change timestamps and will break in ONVIF mode.

They won't have a reference timestamp metadata set and are not useful
for further processing.

onvifmetadataparse: Push buffers from a separate source pad task to guarantee latency and generally improve correctness

We use GStreamer image to build the documentation as it is the simplest
way to ensure that we are testing against the same thing as GStreamer
itself.

followup to 23c07d3c

When we trigger an update pipeline, instead of build-jobs we
replace them update-* jobs. In order for gitlab to not complain
add both of the jobs in the needs array, but mark them as optional
so either of them can be used depending on the usecase.

rajneeshksoni authored Sep 09, 2022 and Sebastian Dröge committed Sep 13, 2022

When retry-duration is less than request_timeout, only 1 try
is attempted.