Issue a warning on gnu void pointer arithmetic

[wrongnull]

Closes #759
In this request I implemented the Type.isPtrTo function but only for the .pointer type specifier. I'm not sure if it is right approach to do that. Maybe other specifiers should be considered as well. I'm open for the conversation.

[ehaas]

It's not a bad idea but we already have a specialized function for it; you can just use ty.isVoidStar()

[wrongnull]

It's not a bad idea but we already have a specialized function for it; you can just use ty.isVoidStar()

I agree. Implementation of such a function is out of this pull requests scope

[ehaas]

The last condition here (p.comp.diagnostics.options.@"gnu-pointer-arith" == .warning or p.comp.diagnostics.options.pedantic == .warning) is not needed - that is all checked in Diagnostics.add via Diagnostics.tagKind

[wrongnull]

done

[ehaas]

We'll also need a check in the .add branch of the switch where the current check is, to handle things like ptr + 1 or 1 + ptr. In the .sub branch, I just realized we only need to check for a.ty.isVoidStar() because you can only subtract a pointer from a same-type pointer. So there's no valid subtraction where b is void * and a is not. But for addition either one could be void *.

One final area to add the checks would be for .plus_plus and .minus_minus in the suffixExpr and unExpr functions to handle the case of ptr++/ptr--/++ptr/--ptr when ptr is void *.

[wrongnull]

We'll also need a check in the .add branch of the switch where the current check is, to handle things like ptr + 1 or 1 + ptr. In the .sub branch, I just realized we only need to check for a.ty.isVoidStar() because you can only subtract a pointer from a same-type pointer. So there's no valid subtraction where b is void * and a is not. But for addition either one could be void *.

One final area to add the checks would be for .plus_plus and .minus_minus in the suffixExpr and unExpr functions to handle the case of ptr++/ptr--/++ptr/--ptr when ptr is void *.

done