lake build # fails
lake env lake build # fails, replaying the error from the previous build
lake clean
lake env lake build # succeeds, because the build cared about one of the environment variables set by `lake env`
I'm not sure what the desired behaviour is here. Perhaps lake should include the values of lean-relevant environment variables, and if these change trigger a rebuild?
A related case is that there are Mathlib tactics that notice the CI enviroment variable, to do additional sanity checks during CI. However CI=true lake build locally is not reliable due to reuse.
I recently encountered a case where:
I'm not sure what the desired behaviour is here. Perhaps
lake
should include the values oflean
-relevant environment variables, and if these change trigger a rebuild?