Is the input for Huber error or error squared?

[Chen-Kai-Xin]

I realize that for most "robustify" functions, the input is "double e2", but for Huber it is "double e". was the definition different for Huber?

[RainerKuemmerle]

Just in the impl the variable is called e instead of e2. For all kernel implementations, the same value is passed.