If we include the definition of SMatrix after MathFunctions.h (which
relies on HelperOps.h (to forward declare SMatrix) we cannot see the
default arguments.

Patch by Lorenzo Moneta and me!

If we have dependencies which we expect to build (like in this case)
we should use DEPENDENCIES.

LIBRARIES is a terrible misnomer and we probably should
fix that.

This patch teaches rootcling to build the pcms in order and
particularly, ROOTWebDisplay.pcm before the WebGui6.pcm

Use the mathcore macro defined in Math/Error.h for error messages. This fixes a compilation problem observed on mac 10.14

ready to be pasted to the ROOT prompt as well as the expected result.

This patch reduces the duplicate content in the GenVector.pcm thus reducing
the amount of decl merging.

This enables clad-ification of these functions.

Fixes: ROOT-9751

This is part of PR #2820. Fixes Windows!

This patch transforms the pattern:
if (cond){
  return a;
}
else {
  return b;
}

into the equivallent:

if (cond)
  return a;
return b;

The function bodies are simple enough and reasonable compiler should be able
to inline them. The second benefit is that if they are in the header file
they will be available to clad and it can compute automatically derivatives
and gradients.

It allows to remove hacks in LinkDef.h, which stop working after removing globbing of headers

instead of vector<bool, RadoptAllocator<bool>>.
It makes no sense to adopt memory in case T is bool since vector<T> becomes
an entirely different class.
As a corollary, the types data_t and const_data_t are defined to describe the
return type of RVec<T>::data(). If T is not a bool, these are T* and const T*
respectively. Otherwise they are both void, a solution as elegant as the one
adopted by the STL for the vector<bool> specialisation.

despite being optional, this is required by gcc 4.8 and gcc 4.9