video-transition: Make smpte props children properties

And deprecate old style accessors.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/207>

ci: include template from gst-ci master branch again

Back to development

Release 1.18.0

demux: Fixate documentation caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/210>

complex_effect_bin_desc: Regenerate expectation for compositor change

Part of: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/796

tests: Fix running tests fully uninstalled

Fixes https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/issues/118

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/206>

Release 1.17.90

ges:source: Handle missing elements in converters

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

smart-mixer: Move the videoconvert to after the mixer

So that it tries to negotiate with alpha and the alpha channel is
dropped as late as possible in the pipeline.

The compositor is able to do video conversion internally in any case
so having a videoconvert before it is useless.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

transition: Enhance name of the elements

Making it simpler to debug

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

source: Handle missing elements in converter

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

video-source: Stop giving useless name to frame positioner

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

transition: Better document the way alpha is computed for transitions

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

smart-mixer: Use the new 'samples-selected' signal to handle queuing in aggregator pads

Since aggregator introduced queueing in its sinkpads the way we set
properties on the pads is incorrect as it doesn't take it into account.
This fixes the issue by using the newly introduced `samples-selected`
signal in aggregator to set the properties right before the compositing
is done.

Also require the compositor we use to be an aggregator.

And add a validate test for it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

ges:validate: Allow setting keyframes using the clips directly

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/204>

ges-source: Ensure that we output stream with segments in time

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

pipeline: Restrict the presence only if the user didn't explicitly provided one

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

timeline: Add a simplified version of track selection signal

Most user do not need to select several tracks for a single
TrackElement and this signal is not binding friendly so
this is adding a simpler, more user and binding friendly version

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

uri-source: Respect user stream selection

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

launch: Also print the position when disabling validate

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

launch: Print more useful information to stdout

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

build: Add an option to disable examples

And make it yield as in other modules

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

launcher: Re activate smart rendering support

Trying to get the best encoding profile for smart rendering when
the user didn't specify anything.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

ges: Fix smart rendering

Smart rendering has been broken since, mostly forever, but some code
was there pretending it was supported... let's try to stop pretending.

We now keep track of the smart rendering state in the timeline, track
and sources to be able to:

 * tell decodebin to stop plugging more (decoding elements) as soon as
   downstream supports the format.

 * avoid plugging converters after the source element when smart
   rendering.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

validate: Pipe debug output to a file when discovering scenarios

Otherwise `gst-validate-launcher` can get veeery noisy

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

nle: Minor debug enhancement

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

tests: Mark audio identity as audio

Otherwise GES fallbacks to video...

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

pipeline: stop setting the track caps

Stop setting the track 'caps' property. The previous code could
overwrite a users own setting of the caps for video and audio caps.
Moreover, the 'caps' property is listed as construct only, and users
will likely expect it to stay the same after a track has been added to a
timeline.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

launcher: Delay setting rendering setting to right before rendering

So that user settings have been applied to the timeline taking into
account any `validatetest` arguments

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

source: Refactor the way we plug converter elements

Paving the way to skipping converters when rendering smartly

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

pipeline: Do not name urisink as `urisink` as it is useless

And actually harmful in case you are debugging several pipelines.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

pipeline: Remove urisink from timeline instead of unrefing it

Doing what was suggested in the FIXME and avoiding to unref
something it while we do not actually own it ourself.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

pipeline: Discard encoding profiles that don't match any track

Otherwise we get a 'not linked' error and we should just help
the user as we can here.

If the user adds a new track, he should set a new encoding profile
anyway.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

uri*source: Factor out common logic into a GESUriSource private data

The two classes are *very* close but have different hierarchy so this
introduces a new GESUriSource structure that is used as private
structure by both subclasses and makes most of the logic shared this
way.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

*uri-source: Call free from the object ->finalize not ->dispose

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/198>

meson: install bash completion helper for ges-launch-1.0

Fixes #77

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/203>

meson: add 'tools' and 'examples' options

To optionally disable build of those.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/203>

Fix retrieving asset metadata on project reload.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/202>

tests: Add test for ges_clip_get_top_effect_index

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/201>

tests: clip: fix test_rate_effects_duration_limit

Fix this assertion:
g_value_copy: assertion 'g_value_type_compatible (G_VALUE_TYPE (src_value), G_VALUE_TYPE (dest_value))' failed

'tempo' is a float, not a double.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/199>

build: Add version.h to the headers list

So it is properly installed and the gir contains the required information

Fixes https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/issues/75

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/197>

pitivi-formatter: Also skip the class

meson: set release date from .doap file for releases

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/196>

title: Make deprecated symbols visible API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/195>

Back to development

Release 1.17.2

docs: fix links

plugins_cache: add base classes

meson: mark plugins cache target as always stale

docs: mark more types as plugin API

doc: Stop documenting properties from parents

smart-video-mixer: Don't call gst_ghost_pad_construct() anymore

It's deprecated, unneeded and doesn't do anything anymore.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/192>

Back to development

Release 1.17.1

ges: Add all missing Since markers from 1.16 onward

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/191>

asset: Do not try to update proxies when we are in a proxying loop

This is a regression introduced in
c12b84788d197c714ec32653e2b751079e377c46, this commit simply brings back
the previous behavior.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/issues/113

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/185>

ges: Refactor the way we plug converters in effects

Stopping to do it at the bin description level but properly
plugging them where they are needed and cleanly ghosting the pads
where it makes most sense.

This introduces support for GES to request pads on the most upstream
element in case no static pad can be ghosted.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/187>

structured-interface: Add support for setting effects inpoint

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/187>

track-element: Make set_has_internal_source return a boolean

Telling the user if it is legal to have an internal source in that
particular GESTrackElement.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/187>

pipeline: doc: Add a note about trying to render before setting rendering settings

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/189>

uri-clip: Add a warning about synchronous uri discovery

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/189>

docs: Update plugins cache

docs: Update plugins cache

tests: enforce I420 format

Tests are assuming video is I420 with a specific chroma and colorimetry
but were not actually enforcing it.
Fixes needed as I420 will no longer be the first video format, see
gst-plugins-base!689

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/184>

track, composition: mark stream id properties as DOC_SHOW_DEFAULT

and update plugins cache

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/182>

doc: Require hotdoc >= 0.11.0

docs: Update gst_plugins_cache.json

add missing Since annotations on new API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/181>

formatter: Do not dereference NULL pointer

CID 1461701

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/180>

xml-formatter: Add an GST_ERROR when setting control sources fails

CID 1463853

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/180>

validate: Wait for state change to consider commit as done

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/178>

validate: Stop always muting

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/178>

formatter: Fix saving/loading project with clip speed rate control

We need to ensure that clips duration is set after time effects are
added and we now need to serialize effects inpoints and max duration.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

docs: add some notes on Time in GES

These notes cover time coordinates in GES, time effects, time
translations.

It also goes into why keyframes will not work with non-linear time
effects.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

uri-clip: don't assume duration needs to stay the same

ges_uri_clip_asset_get_duration does not tell us what the duration in
the timeline needs to be. Especially when we have time effects, or
effects with finite max-durations. So we should no longer expect the
duration to stay the same when replacing assets. Instead, we just check
that the new max-duration would be compatible with the current in-point
(which was not checked before), and the clip would not be totally
overlapped if its duration-limit changes.

This is based on the assumption that each source is replaced one-to-one
in its track. If a source is replaced with nothing in the same track,
this check may be a little too strong (but still mostly weaker than
before). However, problems could occur if track selection does
something unexpected, such as placing the new source in a track not
previously occupied.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

clip: provide an example of using time effects

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

track-element: use out-point for updating control bindings

The out-point, which is an internal time, is used instead of the
duration for determining the control binding value at the end of the
element.

Also, allow the user to switch off the auto-clamping of control sources
if they are not desired. And allow them to clamp specific control sources
individually.

Also, fix a lot of memory leaks related to control sources. In
particular, releasing the extra ref gained by source in
g_object_get (binding, "control-source", &source, NULL);

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

clip: test for layer in group

Make sure the layer exists before we try to remove the grouped clips
from it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

timeline-tree: make sure the layer priority refers to an existing layer

If a layer priority sits between the priorities of two layers in the
timeline, i.e. it references a gap in the timeline's layers, then
ges_timeline_append_layer will never fill this gap and create the
desired layer, so the edit in timeline-tree would loop forever. So a
check was added to avoid this.

This would be a usage error, but a user can reasonably end up with a gap
in their layers if they remove a layer from the timeline.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

clip: add method for adding top effects

Unlike ges_container_add, this lets you set the index and will check
that track selection did not fail. This is useful for time effects whose
addition would create an unsupported timeline configuration.

Also can use the clip add error in ges_timeline_add_clip to let the user
know when adding a clip to a layer that its in-point is set larger than
the max-duration of its core children.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

timeline-tree: take time effects into account when trimming

When trimming the start of a clip, we want to set the in-point of its
children such that whatever data was at the timeline time T still
remains at the timeline time T after the trim, where
  T = MAX (prev_start, new_start)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

clip: use time translation for split

The new in-point should be the media position corresponding to the media
position. media_duration_factor is no longer needed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

clip: fix warning when getting duration-limit

The duration-limit case was missing a 'break;' statement.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

clip: add methods to convert between time coordinates

Add methods to convert between the timeline time coordinates and the
internal time coordinates of a track element in a clip, taking time
effects into account.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

effect: Add support for time effects

Allow the user to register a child property of a base effect as a time
property. This can be used by GES to correctly calculate the
duration-limit of a clip when it has time effects on it. The existing
ges_effect_class_register_rate_property is now used to automatically
register such time effects for rate effects.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

errors: added edit errors

Added more errors to GES_ERROR for when edits fail (other than
programming or usage errors). Also promoted some GST messages if they
related to a usage error.

Also added explanation of timeline overlap rules in user docs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/177>

tests: Stop recording gaps in seek_with_stop

We have little control over those as they are generated by streamsynchronizer in a not reproducible way

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

docs: Remove reference to deinterlacing props in title and video test source

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

validate: Fix setting ges properties

And fix typos.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

ges: Move TimeOverlayClip out of GESTestClip

This was complexifying the implementation for very little gain.
Each source type should ideally have its own API.

In that patch we make it so we do not have to subclass anything
but instead use GESAsset to pass information about how the pipeline
should look like.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

nlecomposition: Add stack initialization action after setting our state

Otherwise there is a pretty rare race where we get the
_initialize_stack_func executed leading to the stack set up and
the source pushing buffers before the composition source pad is
activated, and a STREAM_ERROR is reported as we end up pushing a
buffer to a flushing pad.

Thanks rr chaos mode for showing that improbable race

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

timeline: No thread checking while disposing

While this is not correct, we can't predict from what thread a
GstElement will be disposed as it might still be referenced by
a GstMessage somewhere which is freed by, any thread.

In this specific case we can assume that GES user will already have
let go his timeline reference and we should not avoid assert in that
specific case as it should be safe to let the timeline be destroyed
at that point.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

nle: Use G_PARAM_DEPRECATED for media-duration-factor

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/175>

ges: Ensure that assets are added to project before adding clip to timeline

It is the right ordering and in Pitivi we set the project size
when adding the first (relevant) asset, meaning that our code to
reposition clips would kick in (in the unit tests) if we do not respect
that ordering.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/176>

track-element: Add is_core method to API

Open up the method to the user, since they may need the information.
Also added more documentation on what a core track element is to a clip
and how they are treated.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/169>

asset: unref requested assets

Prevent a few memory leaks in the tests.

Also mark ges_project_save as transfer full for the formatter asset.

Also make sure that ges_project_request_sync is transfer full on the
returned asset.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/issues/104

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/169>

clip: enforce duration-limit

Prevent setting of properties or that of children, if the clip would not
be able to set the corresponding duration if the duration-limit would
drop below the currently set duration.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/169>

clip: make sure core child is active for non-core in same track

Each active non-core child must have a corresponding active core child
in the same track. Therefore, if we de-activate a core child, we also
need to de-activate all the non-core children in the same track.
Similarly, if we activate a non-core child, we need to activate the
corresponding core child as well.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/169>

clip: be more robust in handling priority

Make less assumptions about the priority of effects and core elements so
that the code would still work if the priority of an element was set
directly. In particular, the index of a top effect will always be its
position in the effect ordering.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/169>