What I normally do is: only storing the byte position during the parsing itself, and then extracting the line + column when/if displaying an error.

This allows the error report mechanism to be decoupled, and the hot path of the parser has a bit less code to manage.