When dealing with key-unit trick mode downloads, the goal is to
provide the best "Quality of Experience". This is achieved by:
1) maximizing the number of frames displayed per second
2) avoiding "stalling" as much as possible (i.e. not downloading and
decoding frames fast enough)
This implementation achives this by:
1) Knowing very precisely the current keyframe being download (i.e
more accurate than at the fragment level which might contain more
than one keyfram). This is the new "actual_position" variable
introduced by this commit
2) Knowing the position of downstream (provided by QoS and stored
in the adaptivedemuxstream qos_earliest_time variable)
3) Knowing how long it takes to request and fully download a keyframe
(the average_download_time variable)
Taking those 3 variables into account, whenever a keyframe has been
pushed downstream we calculate a "target time" (target_time variable)
which is the ideal next keyframe time to request so that:
1) It will be requested/downloaded/demuxed/decoded in time to be
displayed without being too late
2) It will not be too far ahead that it would cause too few frames
per second to be displayed.
How far ahead we will request is inversily proportional to how close
the actual position (actual_position) is from the downstream
position (qos_earliest_time). The more is buffered between the source
and the sink, the "closer" the target time will be, and therefore
the more frames per seconds will be displayed (up to the limit
of keyframes_per_second * absolute_rate).
When extracting an aux buffer from an MJPG carrier, at
*least* put the original timestamp on it, even if we
fail to apply any other timestamp (which we always do
at the moment, because the timestamp calculating code
was never finished). Apply a DTS using the camera
supplied delay value as well, assuming that there's
no re-ordering going on (there isn't in the C920,
which is really the only extant camera doing this
stuff) and a warning if that turns out not to be true.
If a manifest has non-zero presentation time offset
(i.e., earliest presentation time specified by sidx box is not zero),
the initial sidx position shouldn't be zero. Since we cannot define
exact sidx position until parsing sidx box, set the value to unknown.
https://bugzilla.gnome.org/show_bug.cgi?id=782693
This embeds the muxer inside the sink and accepts elementary streams
while the old HLS sink required the muxer outside. Apart from that the
interface is the same as before.
Currently only mpegtsmux is supported, but support for other muxers is
just a matter of adding a property.
The advantage of the new sink is that it reduces complexity a lot and
properly handles pre-encoded streams with appropriately spaced
keyframes.
https://bugzilla.gnome.org/show_bug.cgi?id=781496
This is basically a frame counter provided by the driver and it's
advancing at the speed of the HDMI/SDI input. Having this available on
each buffer allows to know what constant-framerate-based timestamp each
frame is corresponding to and can be used e.g. to write out files
accordingly without having the local pipeline clock timestamps used.
https://bugzilla.gnome.org/show_bug.cgi?id=779213
The main advantage is that our sleeps can be interrupted in case of
an src_reset(). Earlier, we would need to wait for a read to complete
before we could do a reset, which could take a long time.
https://bugzilla.gnome.org/show_bug.cgi?id=781249
The audio packet times can be completely unrelated to the video stream
time, depending on the card. While this looks like a bug in the driver,
just always using the video stream time (which is correct) works as a
workaround for now.
This patch bumps the required meson to 0.40.1 as gstreamer core just
did, and cleanup some code to use a feature from 0.37 that allow
specifying version range when checking dependency.
https://bugzilla.gnome.org/show_bug.cgi?id=780654
Otherwise fall back to glDrawBuffers. Also check if glReadBuffer exists
before using it.
glDrawBuffer does not exist for GLES, only glDrawBuffers does.
https://bugzilla.gnome.org/show_bug.cgi?id=782376
This commit fixes the following assumptions with live seeking:
1) start was always valid and of type GST_SEEK_TYPE_SET
2) direction was always forward
3) stop should be offsetted when handling non-accurate seeks before
the range start position.
In order to handle more live seeking use-cases (including reverse playback),
only do non-accurate start/stop value clamping for GST_SEEK_TYPE_SET values.
Also add a bit more debugging lines for issues
https://bugzilla.gnome.org/show_bug.cgi?id=782330
When dealing with live streams, we can't rely on GstSegment calculation
since it uses the segment duration to calculate the absolute values.
But since we are dealing with live *and* we know the ranges, we can
compute the absolute seeking values using the range stop (i.e. "now")
as the END position.
Allows seeking back to "live" by using start_type:GST_SEEK_TYPE_END
and start:0
https://bugzilla.gnome.org/show_bug.cgi?id=782228
We were only ignoring the listed msvc warnings for C language
files and not C++. This was working by the coincidence that we did
not have any instances of these warnings in C++ files. Lately the
build of decklink has been fixed on windows, and it has an
instance of one of these warnings in a C++ file.
https://bugzilla.gnome.org/show_bug.cgi?id=782345
Earlier, the plugin was ignoring those settings and blindly setting
buffer-time to 2 seconds and latency-time to 200ms, which forced all
pipelines to have a minimum latency of 200ms + sink latency.
The values of segsize and segtotal were also not derived correctly.
Now we obey these values, and you can get close to the previous
behaviour by setting buffer-time and latency-time manually. Note that
they are set in microseconds.
As a consequence, when we haven't received enough data from the
device, we now sleep for a time proportional to the data remaining.
However, Directsound is a deprecated API so it maintains its own
software ringbuffer which updates at arbitrary intervals. Hence we
might have to wait a full segsize to get the last 10% of data. To
avoid tight loops, we clamp our sleep floor at 10ms.
In my testing, this keeps the wakeups not-too-high (proportional to
the latency-time set on the source). Further improvements should be
made by fixing the WASAPI audio source plugin instead of this.
Directsound is deprecated and as the comments explain, it is
impossible to get low latency, decent quality, or good performance
from it.
Based on a patch by Sebastian Dröge <sebastian@centricular.com>
https://bugzilla.gnome.org/show_bug.cgi?id=781249
