adjust df_object constraint

[ab9rf]

some (most, actually) get methods return a pointer to a const instance of a subclass of type_identity which isn't convertible to a pointer of non-const type_identity. adding const fixes this. a pointer to non-const is always convertible to a pointer of const of the same type or of a compatible type

fixes #5450

the reason why the pointer to vector glossed as a boolean is that pointers have an implicit conversion to bool, and so when the type inference for the template failed, it fell back to the next best overload, which was the one for bool

[lethosor]

Is this something we could catch in a unit test? I guess it would require a C++ function that takes an affected type...

[ab9rf]

Is this something we could catch in a unit test? I guess it would require a C++ function that takes an affected type...

I should probably add some static_asserts, since that's basically how I found it (along with single stepping).

This particular one is insuring that Lua::Push(lua_State,bool) doesn't get "grabby", since that overload will capture any type that has an implicit conversion to bool (and lots of things have implicit conversions to bool) and doesn't get matched
without a conversion chain.

More generally, I would like to see some C++ unit testing but we don't currently have a good framework for that. Suggestions welcome.

[myk002]

C++ unit testing

We do have cpp unit tests set up. There's an example unit test for miscutils

https://github.com/DFHack/dfhack/blob/develop/library/MiscUtils.test.cpp