The simulation of compressible flows experienced in the ’90s a small revolution with the arising of new algorithms able to compute flows through (or around) any kind of shape. This was due to new numerical algorithms and to new mesh generation algorithms. For both type, the main innovation was related to unstructured meshes, and the way to do it first was to rely on tetrahedrisations. The “any kind of shape” slogan has been progressively completed by the adaptation to any kind of flow, for flows in moving meshes and for automatic mesh adaptation, appearing as an important issue to address in order to improve the expected benefits from a numerical simulation.