This can happen when seeking via http.

This is interlocked with the control queue as it most probably should.

Also make sure that no new controls are pushed after input_Stop().

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>

And also with dxgidebug.h to show leaks
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>

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

The input is already stopped if it has reached end or error.

After each and every MainLoop() call is a call to End(), which
inconditionally sets the input state to END_S.

```--
  CC       access_output/file.lo
access_output/file.c: In function 'Open':
access_output/file.c:288:9: error: implicit declaration of function 'setmode' [-Werror=implicit-function-declaration]
         setmode (STDOUT_FILENO, O_BINARY);
         ^
cc1.exe: some warnings being treated as errors
make.exe[4]: *** [access_output/file.lo] Error 1
```

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

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

Use signed variables for calculating the buffers to be sent to the component
as negative values may occur.
Signed-off-by: Julian Scheel <julian@jusst.de>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Save some memory by allocating fewer buffers. This is still not optimal, but
close to where we can get with the current implementation.
Signed-off-by: Julian Scheel <julian@jusst.de>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>