This way the command line formatter actually uses an URI and not
an ugly hack where were passing a random string instead of an URI.
This also allows the `gessrc` element to handle timelines described
in its URI meaning that you can now use, for example:
gst-play-1.0 "ges:+test-clip blue d=4.0
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/227>
Instead of having it all handled by the tool, this way we can
set the restriction before clips are added to the timeline,
leading to better behavior in term of video images placement
in the scene.
Without that we would have the clips positioned before setting the
restriction caps which leads to weird behavior for the end users.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/227>
It looks like the _set_rendering_details call is superfluous
in _startup(), as it will get called in run_pipeline.
The problem with calling it before timeline_set_user_options
is that we are going to fail creating a smart profile if
the user selected eg --track-types=video, as the get_smart_profile
method compares the tracks in the asset with those on the timeline.
Reproduce with a video-only clip:
ges-launch-1.0 --track-types=video +clip file://$PWD/jelly.mp4 \
inpoint=15.0 -o foo.mp4 --smart-rendering
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/228>
Force a string type for structure values obtained through parsing a
serialized timeline by inserting a (string) specifier after a '=',
rather than relying on gst_structure_from_string guessing the type.
As such, the functions that extract clocktimes and properties are
modified to accept string value types.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/123>
The issue is that we rely on `decodebin::autoplug-select` to `SKIP`
unwanted pads, that signal was first provided to select factories during
autoplugin, not totally thought to avoid exposing pads. For streams
muxed directly in raw, decodebin has nothing to plug after the demuxer
and the pad is exposed right away, meaning that we do not have any
chance to avoid that pad to be exposed. This patch takes that limitation
into account and checks the stream ID of the pads exposed by decodebin
before exposing them itself, so we end up using the right pad even if
more are uselessly exposed by decodebin.
Fixes https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/issues/126
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/222>
Static build fails since version 1.17.1 and
1e488d4311
on:
FAILED: tools/ges-launch-1.0
/srv/storage/autobuild/run/instance-2/output-1/host/bin/arm-linux-gcc -o tools/ges-launch-1.0 tools/ges-launch-1.0.p/ges-validate.c.o tools/ges-launch-1.0.p/ges-launch.c.o tools/ges-launch-1.0.p/ges-launcher.c.o tools/ges-launch-1.0.p/utils.c.o -Wl,--as-needed -Wl,--no-undefined -Wl,-O1 -Wl,-Bsymbolic-functions -static -Wl,--start-group ges/libges-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgstreamer-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgobject-2.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libglib-2.0.a -pthread /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libpcre.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libffi.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgmodule-2.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgstbase-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgstvideo-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgstpbutils-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgstaudio-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libz.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgsttag-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgstcontroller-1.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgio-2.0.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libmount.a /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libblkid.a -lm -Wl,--end-group
/srv/storage/autobuild/run/instance-2/output-1/host/opt/ext-toolchain/bin/../lib/gcc/arm-buildroot-linux-uclibcgnueabi/8.3.0/../../../../arm-buildroot-linux-uclibcgnueabi/bin/ld: /srv/storage/autobuild/run/instance-2/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libc.a(err.os): in function `warn':
err.c:(.text+0x1d8): multiple definition of `warn'; tools/ges-launch-1.0.p/utils.c.o:utils.c:(.text+0x9bc): first defined here
So rename warn function to ges_warn
Also prefix ok, print and printerr function by ges_ for consistancy and
run gst-indent on tools/ges-launcher.c
Fixes:
- http://autobuild.buildroot.org/results/2a528a1185644f5b23d26eb3f2b342e99aa1e493
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/216>
Actually check the version constraint when looking for the glib
dependency.
The version check will make meson use the fallback dependency when the
one from the system is not recent enough, and eventually make the build
succeed even on some older systems like Ubuntu 16.04.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/merge_requests/212>
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>
