This article provides a comprehensive view on the existing knowledge related to adaptation to soil water deficit in maritime pine, a conifer species widely planted in the southwestern Europe. It synthesizes discoveries made in ecophysiology, quantitative and population genetics as well as in genomics, combining several layers of information at the genotypic, phenotypic and environmental levels. Particular focus is given to two major traits: water-use efficiency (WUE) and cavitation resistance (CR). The former is related the maintenance of productivity during periods of lower soil water availability, whereas the latter is tightly linked to survival during severe drought. The development of high throughput phenotypic technologies have made it possible to estimate genetic and environmental variance components of these key traits, providing clues about their suitability for breeding and the evolutionary forces that have shaped their variability. Both CR and WUE were screened in different ecotypes as well as in the Aquitaine breeding population, the main genetic resource of the most advanced maritime pine breeding program in Europe. While the unexpectedly low level of variation of CR within and between natural populations will most likely hamper its use in breeding application, for WUE the medium heritability, absence of unfavorable phenotypic and genetic correlations with diameter growth, as well as the high inter-site correlation and weak genotype-by-environment interaction indicates that artificial selection could be applied for this trait without unfavorable consequences for radial growth, at least within the Aquitaine provenance. On the other hand, recent advances in sequencing and genotyping technologies have contributed to reveal the genetic architecture (i.e. number, location and effect of quantitative trait loci) of these two traits. In combination with ultra-dense genetic linkage map and functional genomics approaches, these findings will contribute to identify positional and expressional candidate genes that should be validated by association genetics and eventually introduced in genomic prediction models to make such knowledge useful to improve tree breeding.