Say, the volume is $V$ and the roll is a cylinder of height $h$ and radius $r$. Then $\pi r^2h=V $ hence $h = V/(\pi r^2)$. As the radius $r$ approaches $0$, the height $h$ grows indefinitely. One expresses this as $$ \lim_{r\to 0} V/(\pi r^2) = \infty $$ The volume does remain constant in the process.
The apparent contradiction arises when you imagine that $r$ actually _becomes_ $0$ in the process, and then conclude that the volume suddenly drops from $V$ to $0$. But $r$ never becomes $0$ in reality. It can be arbitrarily small, but not zero.
To speculate about the volume of a "cylinder" with zero radius and infinite height is about as productive as to debate what would happen if an unstoppable object collided with an immovable one.