Let $A\subset E$ and $B\subset E$ two convexe sets, non empty and disjoint. Suppose $A$ open. There is a closed hyperplan that separate $A$ and $B$.
In the book "Analyse fonctionnelle : théorie et application" (sorry it's in french), page 6 (demonstration théorème I.6) they do as follow : Set $C=A-B$. In particular $C$ is convexe (easy to prove), open (since $C=\bigcup_{x\in B} A-y$) and $0\notin C$ (because $A\cap B=\emptyset$.)
Question : What's the thing with this $C$ ? Since $A$ and $B$ is disjoint, $C=A-B=A$. So nothing to prove about the convexity of $C$ and the fact that $C$ is open... I don't understand what they try to do. Did they made a mistake somewhere ?