Skip to content

Strip every launcher's variables before the serial reference child (#675) #681

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lmoresi merged 1 commit into from
Sep 3, 2026
Merged

Strip every launcher's variables before the serial reference child (#675) #681

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion tests/parallel/serial_reference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,7 +42,13 @@
import underworld3 as uw

# Launcher variables that would make a spawned singleton try to join the parent job.
_MPI_ENV_PREFIXES = ("OMPI_", "PMIX_", "PMI_", "MPI_", "HYDRA_", "I_MPI_", "SLURM_")
# Every implementation we run under is named here, because a list that covers only
# the one on your desk passes locally and fails on the other one: a copy of this
# naming just the Open MPI family passed under Open MPI and aborted the child's
# MPI_Init under MPICH, whose variables are PMI_* (#675). PRTE_/PRTERUN_/OPAL_ are
# Open MPI 5's runtime, which is a separate family again from OMPI_.
_MPI_ENV_PREFIXES = ("OMPI_", "PMIX_", "PMI_", "MPI_", "HYDRA_", "I_MPI_", "SLURM_",
"PRTE_", "PRTERUN_", "OPAL_")

_CACHE = {}

Expand Down
20 changes: 11 additions & 9 deletions tests/parallel/test_0855_mesh_smoothing_parallel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,8 @@
import underworld3 as uw
from underworld3.meshing import smooth_mesh_interior

from serial_reference import _MPI_ENV_PREFIXES


pytestmark = [pytest.mark.mpi(min_size=2), pytest.mark.timeout(120)]

Expand Down Expand Up @@ -248,21 +250,21 @@ def test_parallel_matches_serial_bit_identical():
comm = MPI.COMM_WORLD
rank = comm.rank

# 1. Compute the serial reference (rank 0 only) in a clean
# subprocess. We strip MPI/PMIX/PRTE env vars first so the
# subprocess's PETSc doesn't try to attach to this mpirun's
# MPI world (which would deadlock both processes).
# 1. Compute the serial reference (rank 0 only) in a clean subprocess. The
# launcher's variables are stripped first, or the child's MPI_Init tries to
# join THIS mpirun's job and aborts on a descriptor it does not own.
#
# The prefix list is shared with serial_reference rather than restated here:
# a local copy that named only the Open MPI family passed on Open MPI and
# failed under MPICH, whose variables are PMI_* (#675). One list, so a
# launcher that is missing from it is missing everywhere and gets noticed.
ref_path = None
if rank == 0:
tmpdir = tempfile.mkdtemp(prefix="winslow_ref_")
ref_path = os.path.join(tmpdir, "ref.npz")
clean_env = {
k: v for k, v in os.environ.items()
if not (k.startswith("OMPI_")
or k.startswith("PMIX_")
or k.startswith("PRTE_")
or k.startswith("PRTERUN_")
or k == "OPAL_PREFIX")
if not k.startswith(_MPI_ENV_PREFIXES)
}
proc = subprocess.run(
[sys.executable, "-c",
Expand Down
Loading