is it possible for you to enable the installation of a newer version of dolfinx_mpc via conda?
In dolfinx_mpc:v0.10.0 (relies on dolfinx:v0.10.0), I noticed some strange behavior with nested solvers when solving the Stokes equations with periodic boundary conditions. However, in the newer docker images of dolfinx_mpc:v0.10.1 and dolfinx_mpc:v0.10.2 (both rely on dolfinx:v0.10.0.post2), the error seems to be solved (Thank you!).
On the HPC, I am using Conda to install dolfinx_mpc.
It would be great if the newest version could be installed this way! Thank you very much.
PS: If you need, I can also upload a working example.
Have you tried to install dolfinx_mpc within the same command you create the environment and install other fenicsx and petsc libraries? This has recently solved me a problem when I’ve installed in the complex mode.
I can install dolfinx_mpc via conda install -c conda-forge dolfinx_mpc. However, I then get version v0.10.0.
For the version v0.10.1, the release notes are
”This release addresses a crucial bug-fix for the sparsity pattern creation non-square systems, and blocked systems with mpcs on different blocks.”
Perhaps this bug-fix solves the problem I encountered.
I also tried to install dolfinx:v0.10.0 via conda and then manually install dolfinx_mpc:v0.10.2 by cmake -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX -B build-dir cpp/ ninja -j3 install -C build-dir python3 -m pip -v install --config-settings=cmake.build-type="Release" --no-build-isolation ./python -U --verbose
However, I get error messages when creating periodic boundary conditions. In the past, this problem was connected to nanobind, but I tried to use a proper version.