You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
To do downstream testing of a dist, let's say PDL-LinearAlgebra, we need to
build the current source of its upstream first (PDL).
One way to do this is to build PDL from git then build PDL-LinearAlgebra from
git inside of a single image. But once we have Docker images for PDL, it will
be faster to use the Docker image for PDL that is tagged at the HEAD of master.
Though to keep going further downstream, we aren't necessary creating tagged
containers for every step along the way (e.g., a container for PDL →
PDL-LinearAlgebra → PDL-Opt-NonLinear), so we may still need to build from
source for the intermediates.
Or we can forget the whole thing and build from CPAN for everything except the
current repository under test. Which would be a lot fewer independent
variables.
My thinking is that each thing that want to test its downstreams (including PDL itself) would have a given "from" image, build itself, then be used to test its downstreams. So:
pdl: from=perl, downstreams=(a ref to top of tree)
pdl-linearalgebra: from=pdl, downstreams=(a ref to P:LA's bit)
If there's no explicit tree, the ref would be a condition such as "depends-on" then a value.
e.g., PDL-LinearAlgebra.
The text was updated successfully, but these errors were encountered: