Try to find a working videosink, don't hardcode xvimagesink. Also
add some borders to window and give it a title so that it's clear
that this is really a Gtk+ window and not a window created by the
videosink.

The peer closing the socket should cause an EOS, instead of
silently doing nothing.  This changes the behavior to be
more like fdsrc.  Fixes: #610386

Fixes #610268.

Fix mixup of height/width, causing only half the chroma lines to
be copied when outputting buffers.  Fixes: #610329.

Add simple videotestsrc ! xvimagesink examples using gtk and qt. This patch also
adds all boilerplate to configure for using c++. The qt based examples are
optional like their gtk counterparts.

Correct name of included files. Remove files that are not used anymore. Add many
new api entries to their sections.

We don't do this is any other example, this should be done for us in gdk it if
would be needed.

Use the same translated message string for missing core elements as
playbin uses, which is a bit nicer and also indicates that there is
something wrong with the user's GStreamer installation (which arguably
is the case if elements like typefind or queue2 are missing).

Fixes #609314.

... which only needs a bit of refactoring and extracting to support
the minor difference in (i)vorbis interface.

Fixes #609063.

... such as assuming float all over, and base src caps on template caps.

And post a warning in cases where we can still continue to work
or an error when the missing element is fatal.

They're all working and valgrind clean now.

This makes sure that there *always* is a missing plugin message in the bus
before any errors or warning messages.

and not CORE MISSING_PLUGIN.

Especially if no suitable URI source can be found.

Otherwise the ghostpad will still be linked to the peer and there
will still be a reference kept, leading to nothing being unlinked
and destroyed until decodebin2 is finalized.

This fixes reuse of decodebin2 if a raw stream is connected to
its sinkpad.

The only difference between the two unit tests right now is,
that the decodebin2 test resets the element to READY before trying
to reuse it instead of NULL. decodebin2 guarantees to be reusable
without going back to NULL.

See #609252.

From 96dc793 to 44ecce7

It now contains a single working unit test and can be enabled.
The other more useful unit tests still need fixing.

This makes sure that we don't destroy the last reference before the
element gets back to NULL state. Fixes assertion failures if a playbin2
instance is reused but different sinks are automatically chosen because
of different caps.

JUNQ has same semantics as JUNK.