Does some general improvements with the internal sink handling.
1) Do not remove and re-add the ghostpad when changing
internal sink
2) Only instantiate the default sink when changing from NULL
to READY if there is no other available
3) Avoid changing the internal sink if not on NULL state
Fixes#598682
Downgrade a warning message to debug. Remove an
already fixed FIXME and add a note about (not-)using
fpsdisplaysink in autovideosink. Change the created
ghostpad to use the name "sink" as it is advertised in
the pad template.
Use GST_IS_BIN instead of G_OBJECT_TYPE to check if the
internal sink is a bin. Using the later won't work when
the sink is not a bin directly (but inherits from one, like
autovideosink).
Fixes#604280
Follow-up on 4111d6321f, the video
sink(s) used by fpsdisplaysink might not have the sync property. So we
check its existence to avoid warning from g_object_set() at runtime.
Fixes#604280
Reads the new caps added to qtdemux by commit
c917d65e6d
and adds its corresponding atoms.
Also adds support for image/x-jpc as it is the same
as image/x-jp2, except that the buffers need to be
boxed inside a jp2c isom box before muxing. To solve
this the QTPads now have a function that (if
not NULL) is called when a buffer is collected. This
function returns a replacement to the current collected
buffer.
Fixes#598916
Adds the mapping of 'classification' tags to writing of
'clsf' atoms for gppmux.
Based on a patch by: Lasse Laukkanen <ext-lasse.2.laukkanen@nokia.com>
Exposes the internally used sink as video-sink property and
makes the default one to be autovideosink instead of
the hardcoded xvimagesink
Fixes#604280
Because of an allocated priv (GstRTPMuxPadPrivate), the element will
leak memory if not gst_rtp_mux_release_pad() is called. This would
previously only happen if release_request_pad() was called explicitly,
somthing that should not be neccesary.
Fixes#604099
In particular, consider DISCONT == !sync, and allow subclass to query
sync state, as it may want to perform additional checks depending
on whether sync was achieved earlier on.
Also arrange for subclass to query whether leftover data is being drained.
In particular, (optionally) provide baseparse with a notion of frames per second
(and therefore also frame duration) and have it track frame and byte counts.
This way, subclass can provide baseparse with fps and have it provide default
buffer time metadata and conversions, though subclass can still install
callbacks to handle such itself.
After all, stream is as-is, and there is little molding to downstream's
taste that can be done. If subclass can and wants to do so, it can
still override as such.
Use the rounding version for improved sync between streams.
Small variations in the duration when muxing might lead to
cumullative wrong timestamping when demuxing.
Fixes#602936
Try to use timestamps even when the stream has out of order
timestamps, only fall back to durations when we detect an
out of order buffer. Improves sync between streams.
Fix order, fix variables that don't exist, like GST_LIBS_LIBS,
use $(LIBM) instead of -lm, and move _LIBS from LDFLAGS to LIBADD.
Spotted by Havard Graff.
Adds support for muxing SVQ3 content. Usually this format
has decoder info that must be passed in the 'seqh' field
in the caps. It is also good to add the gama atom to make
quicktime not crash.
Fixes#587922
Prevents losing sync when remuxing streams with different
start times. The smallest start time is selected as
the base time and all timestamps are subtracted
from it to get the actual time to be used when
muxing and building indexes
Fixes#586848
Do not wrongly add the result of the function to the
pointer to the buffer size. Instead, check the result
to see if the serialization was ok.
Based on a patch by: "Carsten Kroll <car@ximidi.com>"
Fixes#602106
When muxing streams, some can start later than others. qtmux
now handle this by adding an empty edts entry with the
duration of the 'lateness' to the stream's trak.
It tolerates a stream to be up to 0.1s late.
Fixes#586848
Using the end time makes it impossible to replace buffers, which is
a big problem for subtitles that could have very long durations.
Merged from gst-plugins-base, 27034be461.
Scaletempo was missing an update of 'stop' in
new segment parameters when pushing it downstream,
which caused files to end earlier when rate < 1.
Fixes#599903
Based on patch by: Bastian Hecht <hechtb@gmail.com>
It looks at raw audio data and emits messages when DTMF is detected.
The dtmf detector is the same Goertzel implementation used in FreeSwitch
and Asterisk. It is in the public domain.
There is unfortunately no G_*_FORMAT conversion specifier for differences of
pointers in glib, and we can't rely either on all platforms being 64bit.
So let's just cast the difference to a gint and be done with it.
There's not much point in using GST_DEBUG_FUNCPTR with GObject
virtual functions such as get_property, set_propery, finalize and
dispose, since they'll never be used by anyone anyway. Saves a
few bytes and possibly a sixteenth of a polar bear.
Merged from gst-plugins-base, 6f4c1ac583.
Replaced with "GStreamer maintainers
<gstreamer-devel@lists.sourceforge.net>" or just removed,
depending on the number of other authors.
Merged from gst-plugins-base, 0e9bc5125a.
Set the output caps on the srcpad before pushing the buffer because else core
will do a rather expensive check to see if we can actually accept those caps on
the srcpad.
Merged from gst-plugins-base, bdfb4b46d7.
Install a custom acceptcaps function instead of using the default expensive
check. We accept whatever downstream accepts so we pass along the acceptcaps
call to the downstream peer.
Merged from gst-plugins-base, 5b72f2adf9.
Clarify the ownership of the internal plugin feature list by making
a copy of any passed list. Avoids crashes when freeing a passed list,
or leaks caused by not freeing any internally built list.
Also remove GST_PLUGINS_BASE_LIBS from LIBADD since we don't
need to link against any of the -base libs (we just use a define
from the gstaudio headers).
When sending new-segment to a stream, ensure that there is either a valid
PCR, or else wait until there's a PTS on the stream (dropping packets if
needed) in order to avoid generating an invlaid new-segments event.
https://bugzilla.gnome.org/show_bug.cgi?id=595161
g_convert seems to add a single null terminating byte to
the end of the string, even when the output is UTF16, we
force the second 0 byte when copying to the output buffer.
This issue was causing random crashes because it was
assumed that the string resulting from g_convert had
2 extra bytes, but it has only one.
Add the 'initial-identity' property, which inserts identity for
at startup for event passing, and replaces it with a new child
when the first buffer (and caps) actually arrives.
https://bugzilla.gnome.org/show_bug.cgi?id=599469
Keep track of the chunk durations to be able to add 3gr6
brand if it is a faststart file and the longest chunk is
smaller than a sec. Implemented according to 3gpp
TS 26.244 v6.4.0 (2005-09)
Fixes#584361
In faststart mode, there is no need to send the ftyp
right at the beginning of the stream. Waiting and sending it
only later (when the moov atom is ready to be sent) provides
us with more information about the stream and we can better
select the compatible brands.
Align element initialisation. This should be re-thought, g_object_new zeros things already.
Harmonize the element getters for the src/sinks to return what we actualy use.
This uses same approach like in playbin, namely checking for user defined
element, auto{audio,video}{sink,src} and finally DEFAULT_{AUDIO,VIDEO}{SRC,SINK}
defines from config.h.
gst_pad_set_caps on the internal source pad always succeeds, because
caps propagate to the peer with buffers, not immediately. Using
gst_pad_peer_accept_caps properly checks whether the actual
sub-element can accept caps when they change.
https://bugzilla.gnome.org/show_bug.cgi?id=575568
Camerabin caches photography settings, but it didn't take into account
that scene mode setting may change other settings as well. So, config
needs to be read back from device after scene mode is set.
Camerabin incorrectly used G_GUINT64_CONSTANT macro for setting
"max-size-buffers" and "max-size-bytes" properties in image queue,
even when they aren't 64bit integers.
Camerabin sets itself to READY state during resolution change. This
operation makes output-selector to forget its currently active pad,
so it must be set again after state change.
If an error occurs, application should set pipeline to NULL, and updating
zoom can actually block message handling if video device driver has failed
and video src element supports zooming using photography interface and S_CROP.
Improve (slightly) the interpretation of PGS set-window blocks
to avoid printing warnings about unused bytes when there are multiple
window definitions.
Fix the rendering when we hit the right hand side of the display
area, by resetting to the correct X coordinate, and add some more
guards against bad PGS data.
Using a GstIterator is slow because we have to create/destroy that
iterator every single time.
We just do the threadsafe cookie check and list iteration ourselves.
There's no need for an extra function since all debuggin will require
a MpegTSPacketizer which means that the GType will be created, therefore
move the debug category initialization there.
