qreal is a double on some platforms (x86) and a float on others (e.g. ARM), cast it to fix build errors like these:

src/ui/designer/QGCXYPlot.cc: In member function 'int
XYPlotCurve::appendData(const QPointF&)':
src/ui/designer/QGCXYPlot.cc:41:39: error: no matching function for call
to 'qMin(double&, qreal)'
             xmin = qMin(xmin, data.x());
                                       ^
src/ui/designer/QGCXYPlot.cc:41:39: note: candidate is:
In file included from
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtCore/qnamespace.h:45:0,
                 from
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtCore/qobjectdefs.h:45,
                 from
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtGui/qwindowdefs.h:45,
                 from
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtGui/qwidget.h:46,
                 from
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtGui/qdockwidget.h:45,
                 from
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtGui/QDockWidget:1,
                 from src/ui/designer/QGCXYPlot.cc:1:
/build/v2013.12/build/tmp-angstrom_v2013_12-eglibc/sysroots/sama5d3_xplained/usr/include/qt4/QtCore/qglobal.h:1213:34:
note: template<class T> const T& qMin(const T&, const T&)
 Q_DECL_CONSTEXPR inline const T &qMin(const T &a, const T &b) { return
(a < b) ? a : b; }
                                  ^

See http://www.qtcentre.org/archive/index.php/t-45475.html

 for some
background.

This fixes #625

Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>

Proxy now follows system proxy

This fix was pulled across from current QSerialPort codebase. I looked
into pulling across newer version of full QSerialPort. While doing that
I discovered that the version we are using has already been hacked to
work with us. It was fairly involved to redo those hacks. So decided to
pull individual change across instead since our version is already not
clean. When we switch to Qt5 all this will go away.

qserialport and qextserialport use the pattern of += into
PUBLIC_HEADERS and PRIVATE_HEADERS. This leads to the same header being
included twice. Which in turn causes the rule for moc in the makefile
to be added twice, hence the warning. Using *= instead of += removes
the duplication.

The latest version of qwt does have these warnings fixed. But the
latest version also has removed some deprecated printing functionality
which we are currently using. I gave a try at updating but it looked
fairly involved and over my level of understanding for that part of the
code. So pragmas for now. When someone does upgrade qwt the pragmas
will no longer be needed.

I looked at the latest version of opmapcontrol in OpenPilot. This has
not yet been fixed. So added pragma to silence.

qserialport and qextserialport use the pattern of += into
PUBLIC_HEADERS and PRIVATE_HEADERS. This leads to the same header being
included twice. Which in turn causes the rule for moc in the makefile
to be added twice, hence the warning. Using *= instead of += removes
the duplication.

The latest version of qwt does have these warnings fixed. But the
latest version also has removed some deprecated printing functionality
which we are currently using. I gave a try at updating but it looked
fairly involved and over my level of understanding for that part of the
code. So pragmas for now. When someone does upgrade qwt the pragmas
will no longer be needed.

I looked at the latest version of opmapcontrol in OpenPilot. This has
not yet been fixed. So added pragma to silence.

Replaced by /libs/serialport

Replaced by /libs/serialport.

This removes a race condition between both qwt and qt against math.h on
windows with _USE_MATH_DEFINES set. If math.h comes after qwt or qt
math headers the math defines will be redefined. Really nasty to track
down and ordering problems. So instead I just removed using math
defines on windows and fall back to asin definition of M_PI_2. This
will have some slight perf impact since it calls function instead of
using constant. But I doubt on any of today’s machine it really makes a
difference.

Pulled fixe from latest eigen across. Figured it was safer to patch
than to pull across whole new version of eigen.

This reverts commit fcfb392e.

Start move to single column layout for advanced parameters; Fix
crashing bug in PrimaryFlightDisplay when uas is removed and then
re-added; workaround compiler warning in GeneralBlockPanelKernel,

This reverts commit 233a1716.