Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Close #3471

Just deactivate proper seeking

This fixes the RTP input failing completely.

Jitter computation is, by design, relative to the time of arrival of
packets. Normally, it should be the estimated time of arrival (ETA) of
the missing packet in the sequence, plus a tolerance value computed
from the delay variance (jitter). Unfortunately, we do not know that
ETA since we do not know the RTP timestamp of missing packets.

This reverts commit 9d9ffe9f.

Still could be better to leave selected item as selected after sorting.

Not really nice looking or anything, but I'm not that familiar with lua.
Also not sure if comparing paths/uris is the way, didn't spot any way to
get same id from both of those (or missed something trivial).

- Simplify switching between widgets in central stack: showTab() allows
switching only between two states - video state and playlist state - and
does everything needed according to circumstances

- Do not setMinimumWidth/Height/Size() on main interface, because it
inhibits children sizeHint() evaluation, thus preventing interface
self-adjustment when controls are added/removed or shown/hidden.

- Subclass QStackedWidget so it's minimumSizeHint() matches exactly
the one of the currentWidget(), and not the sum of all children.

- Suggest compact mode implementation (commented out for now):
setSizeConstraint( QLayout::SetFixedSize ) on main interface layout.
This automatically keeps and adjusts fixed size in respect to
sizeHint() of children.

Parental advisory: explicit hack.

If kfilemodule is loaded, bypass run cleanup handlers. Library really
should not use this C feature in any case (IMHO). And I find it
outright idiotic for a (Qt4) plugin to use cleanup handlers.
This is causing so many crash reports that it's not funny anymore:
https://bugs.launchpad.net/ubuntu/+source/vlc/+bug/408719

Pointed-Out-By: Gaëtan Rousseaux

The decompression process needs to "see" the end-of-file. Otherwise
it might get stuck, and then the demux thread will get stuck too.
Closing the write end of the pipe to the compression process fixes
that.

Pointed-out-by: Laurent Aimar

(It could be more efficient, but who cares? pipe overhead is probably
 worse)

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>