C++ Using the same default_random_engine for every new object

I'm a C++ beginner coming from a Java and C# background. I'm trying to use the same default_random_engine and normal_distribution at every creation of a new object. Before I was using a new default_random_engine with a new seed and a new normal_distribution in every constructor. I think that way the normal_distribution doesn't work correctly.

Old code ->

my_object.cpp:

default_random_engine generator;

MyObject() {
    double mean = 1.0;
    double std = 0.5;
    normal_distribution<double> distribution(mean, std);
    QTime time = QTime::currentTime();
    uint milli = (time.hour() * 60 * 60 * 1000) + (time.minute() * 60 * 1000) + (time.second() * 1000) + time.msec();
    generator.seed(milli);
    myValue = distribution(generator);
}

This compiled and the values for myValue were randomly distributed. I just think they didn't match the normal distribution, because I always created a new default_random_engine and normal_distribution and used a new seed.

My new code ->

main.h:

class Main
{
   public:
       static default_random_engine generator;
       static normal_distribution<double> distribution;
};

main.cpp:

default_random_engine generator;
normal_distribution<double> distribution;

int main(int argc, char *argv[]) {
    double mean = 1.0;
    double std = 0.5;
    distribution(mean, std);
    QTime time = QTime::currentTime();
    uint milli = (time.hour() * 60 * 60 * 1000) + (time.minute() * 60 * 1000) + (time.second() * 1000) + time.msec();
    generator.seed(milli);
...
}

my_object.cpp:

default_random_engine generator;
normal_distribution<double> distribution;

MyObject() {
    myValue = distribution(generator);
}

But now I get 10 errors on compile time:

error C2228: left of '.min' must have class/struct/union
error C2780: '_Rty std::_Nrand(_Engine &,long double,_Rty)' : expects 3 arguments - 2 provided
error C2780: '_Rty std::_Nrand(_Engine &,double,_Rty)' : expects 3 arguments - 2 provided
...

What am I doing wrong? Why am I getting the errors? And am I correct that earlier my normal distribution/random generation wasn't correct? Will I be able to produce the wanted normal distribution this way?