Signed-off-by: Pierre d'Herbemont <pdherbemont@free.fr>

Up to now, text control could but be focusable, and this focus was used
to scroll text, should it be too long for full display.

From a user viewpoint, this behavior is often counterintuitive :
 - Most of the time, the control is designed to fit the text anyway, and
   scrolling is not expected.
 - On the other hand, an underlying control may be used to move the window
   (for instance, the title bar), and a text control is expected not to
   interact in this endeavor. Same goes for popupmenu, that is expected
   to work if one happens to right-click on a text control.

This patch adds a 'focus' attribute to the text control, and leave it to the
skins developper to decide which behavior is most desirable on a per-control
basis. focus still defaults to true to maintain skins current behavior.

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

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

Return a 413 error instead of crashing

This caused genuine bugs like mangled bodies and corrupted requests
(usually leading to 501 errors), and possibly leaks of memory buffer
of the size of the Content-Length parameter. This is pretty bad since
crafted requests could easily exhaust the memory and/or trigger a crash.
Apart from the contents of the buffer of the involved request, there
doesn't seem to be any possible memory corruption.

This fixes a regression brought by 7f485998

On skins2(tested on Linux), open and close the preferences dialog leads to a
crash, because the whole qt4 thread terminates unexpectedly.

This is due to setQuitOnLastWindowClosed set to true. As a dialog provider,
qt4 doesn't run any windows, and therefore just closing the preferences
 dialog box (via accept() or reject()) signals the qt4 thread to quit.

8554 is the IANA-registered RTSP alternative port

- setContentMargin is the same on all platforms
- color fixes on Linux

For users who'll want to revert to the old one for whatever reason

 - Gtk+ player
 - wx   player
 - Qt   player
 - DVD ripper using Gtk

The new version, merged into the RTP stream output, is supposed to be
better than the old one, except for two dropped features:

 - non-standard RAW/RAW/UDP and MP2T/H2221/UDP RTSP Transports, and
   x-playNow RTSP header: the only vague specification of this that I
   could find is this 10-year-old draft:
   http://www-rn.informatik.uni-bremen.de/ietf/mmusic/47/id/draft-sheedy-mmusic-rtsp-ext-00.txt
   They provide support for broken Amino STBs.

 - client number limiting: not sure how pertinent that is and if it
   would be the right way to implement/port it

If you think it's a problem to drop these features, speak now.
Ultimately the goal is to get rid of the old code (after the next major
release?)

Closes #568

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

They are only used by the thread and can be freed before the module closes.

If search.h is present but not tdestroy: implement it based on others t* functions
If search.h and tdestroy are not present: implement every t* functions.

Close must also be serialized (one single global lock/wait).

function call didn't match function prototype.