Open jakirkham opened 5 years ago
Main trick is just keeping track of state (aka the things that have been built)
@jakirkham so you envision having several recipes in one feedstock that depend on each other? Or would you somehow want to use multiple outputs of one recipe still?
It would still have multiple outputs.
An example of this would be a feedstock with one output myclib
and another output python-myclib
, which builds against all supported python
versions. Here we would want to build myclib
once and then reuse that when building python-myclib
for each python
version.
Today we rebuild myclib
for every python
version as well.
Azure has been advertising their multistage pipelines over the past few months, it could be interesting for us to explore this feature. In particular this could be really handy with split packages. For example a package that builds a C/C++ library as one package and then build Python bindings to that library. In these cases we might benefit from one job for the C/C++ library and multiple jobs for the Python bindings (dependent on the original single C/C++ job).
cc @mariusvniekerk (in case this is of interest 😉)