Open douglowe opened 2 years ago
I wonder if this issue is caused by the same problem: https://github.com/conda-forge/conda-smithy/issues/1223
I can confirm that
# skip: True # [mpi == "nompi" and double == "yes"]
produces 4 variants when it should skip one.
The following statement also does not work properly:
skip: True # [not (mpi == "nompi" or double == "yes")]
Solution to issue cannot be found in the documentation.
Issue
I have found that
conda smithy rerender
will ignoreskip
statements if they use additive logic.For example, I have a build matrix:
If I use the following
skip
statement I get the correct build (just theopenmpi
andno
build):If I switch to using
and
instead ofor
I get all 4 builds, not just the 3 builds that I should get, because thenompi
andyes
build is still included.This error is impacting this recipe: https://github.com/conda-forge/gromacs-feedstock/pull/3 - because I need additive skip statements in order to avoid excess builds.
This problem is reproducible both using the rerender service on github, and running
conda smithy rerender
locally.Installed packages
Environment info