Computation on custom mesh: PETSc error code is: 73 (Object is in wrong state).

[Nico Schlömer]

MWE:

from dolfin import *

mesh = Mesh('test.xml')

V = FunctionSpace(mesh, 'Lagrange', 1)
u = TrialFunction(V)
v = TestFunction(V)
sol = Function(V)
solver = KrylovSolver('gmres')
solver.set_operator(assemble(u*v*dx))
solver.solve(sol.vector(), assemble(Constant(1)*v*dx))

gives

*** Error:   Unable to successfully call PETSc function 'KSPSolve'.
*** Reason:  PETSc error code is: 73 (Object is in wrong state).
*** Where:   This error was encountered inside /build/dolfin-1R8MPk/dolfin-2016.2.0/dolfin/la/PETScKrylovSolver.cpp.

This uses test.xml.

When using mesh = UnitSquareMesh(20, 20), it all works as expected.

[Nathan Sime]

Your mesh looks bad. There's nothing connected to vertex 0. The matrix is singular.

from dolfin import *

mesh = Mesh('test.xml')

V = FunctionSpace(mesh, 'Lagrange', 1)
u = TrialFunction(V)
v = TestFunction(V)
A = assemble(u*v*dx).array()

bad_dof = 0
d2v = dof_to_vertex_map(V)

# Zero diagonal
print A[bad_dof, bad_dof]

# The bad vertex
bad_vertex = Vertex(mesh, d2v[bad_dof])
print "Bad vertex: " + str(bad_vertex.index())

# Connectivity has no cells
mesh.init()
print [c for c in cells(bad_vertex)]
print [f for f in facets(bad_vertex)]

In the xml

<vertex index="0" x="1.0" y="0.0" z="0.0"/>

is never used.

[Jan Blechta]

Good spotting @nate-sime. In fact, one gets for d2v = dof_to_vertex_map(V) when assertions are on

*** Error:   Unable to complete call to function vertex_to_dof_map().
*** Reason:  Assertion vertex_found failed.

[Jan Blechta]

Feel free to open a new issue about error handling in mesh construction.

[Nico Schlömer]

Thanks everyone!