User defined literals (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2378.pdf)
are already available in gcc 4.7 and clang 3.1 and this results in problems in code
where literal strings are concatenated without any whitespace inbetween.

Use a better approach than revision 1387711

The syntax is different and somewhat ugly as it doesn't seem
possible to manipulate directly the real or imaginary parts
but it is possible to use operator overloading.

This is done for "fun" and to keep in line with evolving C++
standards. Numerical results should be unaffected.

that solves another dozen warnings

XMLPropertyMapEntry's signedness is the root cause of the problem,
but that class is widely used outside, even in binfilter, so changing
it directly is not advisable. Fixing it indirectly suffices to
eliminate almost 300 warnings.

XkbKeycodeToKeysym() has obsoleted it in 2003 anyway

Revision 1440387 changed defaultprocessor3d.hxx to be no longer "delivered".
It is needed in svx by being included indirectly via cutfindprocessor3d.hxx.
Fixing the build breaker by "delivering" the missing header file again.

This change eliminates almost 6000 warnings when building on newer UNX platforms 

The destructor of the Reference class depends on its _pInterface member being set,
but some constructors skip this when an exception was thrown. Always initializing
them properly while avoiding a rethrow is cheap and safe.

Fixed-By: Herbert Dürr, Andre Fischer

setting ant's includeantruntime option makes builds more repeatable

Replace WH2006 PRNG with the well known Mersenne Twister
as implemented in Boost. This actually has some drawbacks
with respect to our previous implementation: we lose the
nifty seeding and instead use the timer which is
non-optimal.

The main reason is the speed: MT is said to be at least
4 times faster.

On the other hand I also moved the WH to the scaddins
module where it is  used by Calc's RANDBETWEEN function.
This way we still have the alternative PRNG and we
drop another use of the unreliable rand from libc. 

The root problem is that the boolean expression "true"
evaluates to -1 rather than +1 in formulas in Writer.

The problem is solved by considering that the type can
be a Boolean (SBxBool in addition to SbxSTRING.

Author:	Mattias Johnnson