On Sat, Jan 5, 2019 at 8:11 AM Olivier Lamy <ol...@apache.org> wrote:
> No way. Apache Infrastructure is a shared infrastructure. There are NO > dedicated machine for any projects. > I guess you mean physical machines here, don't you? I don't think physical machines are what matters, but having an infra that can provide some isolated and dedicated VMs/Containers to projects would be nice and probably would ease maintenance and innovation by making everything less risky. It's the current setup of build infra at Eclipse.org ( https://medium.com/@mikael.barbero/scaling-up-the-continuous-integration-infrastructure-for-eclipse-foundations-projects-6fd60d4dc41d ), and it really seems to be profitable to both project developers and infra team compared to previous approaches (multi-projects CI with shared slaves, project-specific CIs with shared slaves, project dedicated CI and slaves...). But anyway, if Travis can run a pass of build+tests that would catch a good ratio of errors and report them to contributors automatically, it could be far enough. With my (low) understanding of the build infra at Apache, it seems to me that enabling Travis is by far the action that would bring the most immediate and biggest ROI here; much more than trying a revolution at Apache build infra. Cheers,
