Using _foreach_remove on the hashtable, while releasing the lock protecting
that table inside the callback is not a good idea. The hashtable might
then change (a source removed or added) while signals like on_timeout
are being sent.
This solution makes a copy of the table, performs the _foreach without
actually removing any sources, but marks them for removal on a second
iteration with the real list, but this time not letting go of the lock.
Fixes#630452
Holding internal locks while potentially calling out is a source
of deadlocks, and in this case the application might subscribe to the
pad-added signal.
Fixes#630449
If the source has been inactive for some time, we assume that it has
simply changed its transport source address. Hence, there is no true
third-party collision - only a simulated one.
Fixes#630447
Using the symbols for the different Speex modes results
in crashes when using MSVC. Use the library functions to
get the modes instead.
Fixes bug #630378.
For H264, there is an extra header containing the CTS, which is a timestamp
offset that should be applied to the PTS. Parse this value and use it to adjust
the pts.
Fixes#630088
Make the is_dead check more clear and add an option to check for the status of
the stream in addition to the context.
We don't need a stream to get the device_description string.
Fixes#630317
Parses uuid atoms in push mode when they are found, they might
contain xmp tags.
Also does a minor refactoring to put the global tags posting
into a single function instead of repeating it in 3 different
places.
Fixes#629839
xmp packet is placed into a top-level uuid atom for
isom/mp4 variants.
This patch makes qtdemux parse all top-level atoms
in pull-mode before starting to push data, making
it able to find those tags.
https://bugzilla.gnome.org/show_bug.cgi?id=629839
We also need to share the main-loop threads as this owns the context. Thus have
a class wide main-loop thread. From this we create a context per client-name.
Instead of always looking up the context, we keep this with the instance. The
reverse mapping is only needed in pulse singal handlers. This saves a lot of
locking. Also one signal handler becomes simpler as ther eis only one mainloop
to notify.
Now valgind happy - no leaks, no bad reads/writes.
This reverts major parts of commit 69a397c32f.
Fixes#628996
