dolfin team mailing list archive

[Joachim Haga <jobh@xxxxxxxxx>]

> What about requiring that it is either of length zero or of correct
> size. Would that break anything?

I would be happy with this (speaking as a current heavy user of the
Matrix::mult interface, which behaves similarly).

Zero size: resize
Correct global size: Use as-is
Incorrect global size: Error

...and maybe add, for parallel-safety (the backends probably do this already?):
Correct global size, incorrect local size: Fail fatally

If I remember correctly, the local size used to be checked, but that
check was removed since trying to correct it (via resize) may lead to
deadlock. A fatal error is the only option in this case.

-- 
You received this bug notification because you are a member of DOLFIN
Team, which is subscribed to DOLFIN.
https://bugs.launchpad.net/bugs/732543

Title:
  No error when attempting to solve using wrong dimensions

Status in DOLFIN:
  New

Bug description:
  Calling solve with a vector of non-matching dimensions does not
  produce an error. Something mysterious happens instead. Reproduce with
  below code.

  from dolfin import *

  mesh = UnitSquare(2, 2)

  V = FunctionSpace(mesh, "CG", 1)
  Q = FunctionSpace(mesh, "CG", 1)
  W = V * Q

  y = TrialFunction(W)
  w = TestFunction(W)

  a = inner(y, w)*dx
  L = w[0]*dx

  A = assemble(a)
  b = assemble(L)
  y = Function(W)

  info(A)
  info(b)
  u = Function(V)
  info(u.vector())
  solve(A, u.vector(), b)
  info(u.vector())