In this work, a memristor-based spiking-GAN neuromorphic hardware system is proposed. Both the generator and discriminator of GAN are in the form of Spiking Neural Network to improve the computational performance, and the memristor synapse circuit with 1 memristor and 4 transistors is proposed as Computing in Memory to avoid the cost of memory accesses. The reinforcement learning rule is used to train both discriminator and generator networks. Tests on the MNIST and Fashion-MNIST datasets showed the proposed GAN can efficiently generate data samples. The results demonstrate the great potential of this memristor-based spiking-GAN for high-speed energy-efficient data augmentations.