Understanding massive stars is essential for a variety of branches of astronomy including galaxy and star cluster evolution, nucleosynthesis and supernovae, pulsars, and black holes. It has become evident that massive star evolution is very diverse, being sensitive to metallicity, binarity, rotation, and possibly magnetic fields. Although the problem to obtain a good statistical observational database is alleviated by current large spectroscopic surveys, it remains a challenge to model these diverse paths of massive stars toward their violent end stage. I show that the main sequence stage offers the best opportunity to gauge the relevance of the various possible evolutionary scenarios. This also allows sketching the post-main-sequence evolution of massive stars, for which observations of Wolf-Rayet stars give essential clues. Recent supernova discoveries owing to the current boost in transient searches allow tentative mappings of progenitor models with supernova types, including pair-instability supernovae and gamma-ray bursts.