Let $P$ be a finite subset of ${\mathbb Z}$ containing at least three elements. I say that $P$ tiles $\mathbb Z$ if $\mathbb Z$ can be written as a disjoint union of translates of $P$.
Trivially if $P$ is an arithmetic progression of the form $\lbrace t+sk \ | \ 1\leq k \leq M \rbrace$ where $s$ and $M$ are coprime, then $P$ tiles $\mathbb Z$ (because of Euclidean division and the Chinese remainder theorem).
Is the converse true ?
My thoughts : It is easy to see that $P=\lbrace 1,2,4\rbrace$ can never tile $\mathbb Z$. For suppose it did, and suppose $\bigcup_{t\in T}t+P$ is such a tiling. We call the $t+P$ (for $t\in T$) distinguished translates of $P$. Choose any $t_1\in T$. Then $t_1+3$ must be in some distinguished translate of $P$, which can only be $(t_1+2)+P$ or $(t_1+1)+P$ or $(t_1-1)+P$. But all those intersect $t_1+P$, contradiction.