Edge case: fd.dx works but fd.par_loop fails when fd.dx subdomain_id is a single-element tuple

[anilbeycorintis]

fd.dx accepts int or tuple as seen below.

fd.dx(8, domain=mesh)
fd.dx((8,), domain=mesh)

When fd.dx is to be called from within fd.par_loop then the single element tuples fail.

Minimal reproducer:

import firedrake as fd

# Mesh with all cells labeled as subdomain id 8
base = fd.UnitSquareMesh(2, 2)
V0 = fd.FunctionSpace(base, "DG", 0)
mesh = fd.RelabeledMesh(base, [fd.Function(V0).assign(1.0)], [8])

f = fd.Function(fd.FunctionSpace(mesh, "DG", 0))

fd.dx(8, domain=mesh)  # OK
fd.dx((8,), domain=mesh)  # OK

# Works with int
fd.par_loop(
    kernel=("{[i]: 0 <= i < f.dofs}", "f[i] = 1.0"),
    measure=fd.dx(8),
    args={"f": (f, fd.RW)},
    is_loopy_kernel=True,
)

# Fails with single-element tuple
fd.par_loop(
    kernel=("{[i]: 0 <= i < f.dofs}", "f[i] = 1.0"),
    measure=fd.dx((8,)),
    args={"f": (f, fd.RW)},
    is_loopy_kernel=True,
)

[jorgensd]

Should this be moved to firedrake, @dham ? As it is not clear that this relates to FEniCS itself, but the firedrake interpretation of integration measures.

[dham]

@anilbeycorintis can you post the traceback please? That will tell us whether this is failing in UFL or Firedrake. I can't say for sure based on your post, but I suspect @jorgensd is right and this is not a UFL issue.