Fix equality for FieldVectors with different type

Before this commit, == for FieldVectors with different types was falling
back to the AbstractArray ==, resulting in false positives. Now
FieldVectors with different types are always considered different.

NEWS.md

@@ -12,6 +12,14 @@ v0.14.17
  - Fixed some type instabilities PR [#2004](https://github.com/CliMA/ClimaCore.jl/pull/2004)
  - More fixes to higher resolution column cases for the GPU [#1854](https://github.com/CliMA/ClimaCore.jl/pull/1854)
 
+### ![][badge-🐛bugfix] Fix equality for `FieldVector`s with different type
+
+Due to a bug, `==` was not recursively checking `FieldVector`s with different
+types, which resulted in false positives. This is now fixed and `FieldVector`s
+with different types are always considered different.
+
+PR [#2021](https://github.com/CliMA/ClimaCore.jl/pull/2021).
+
 v0.14.16
 -------
 

src/Fields/fieldvector.jl

@@ -487,10 +487,18 @@ end
 
 Recursively compare given fieldvectors via `==`.
 Returns `true` if `x == y` recursively.
+
+FieldVectors with different types are considered different.
 """
 rcompare(x::T, y::T) where {T <: Union{FieldVector, NamedTuple}} =
     _rcompare(true, x, y)
 
-# Define == to call rcompare for two fieldvectors of the same
-# exact type.
-Base.:(==)(x::T, y::T) where {T <: FieldVector} = rcompare(x, y)
+rcompare(x::T, y::T) where {T <: FieldVector} = _rcompare(true, x, y)
+
+rcompare(x::T, y::T) where {T <: NamedTuple} = _rcompare(true, x, y)
+
+# FieldVectors with different types are always different
+rcompare(x::FieldVector, y::FieldVector) = false
+
+# Define == to call rcompare for two fieldvectors
+Base.:(==)(x::FieldVector, y::FieldVector) = rcompare(x, y)