Potential issues with Field copy constructors.

[d7919]

I've just come across a potential issue with the field copy constructors. These just give the new field a reference to the existing data block, which we call checkData on if CHECK > 2.

Consider the following case

Field2D a;
a.allocate();
Field2D b(a);

In the copy constructor call to make b we will checkData on the data block of a. This has been allocated but it has not been initialised so the checkData output is undefined (i.e. it could throw or it could not).

I think the only ways to avoid this undefined behaviour would be to either remove the checkData call in copy constructors or to ensure allocate initialises the data block.

[d7919]

For example, I think this is the cause of the test failure here.

[ZedThree]

Fixed in #1533