Principle of the test
================

The problem is a very simple elastic mechanical problem.

Kinematic boundary conditions are not trivial: there are relationships between several degrees of freedom.

The reference calculation is obtained with solver MUMPS and using ELIM_LAGR =' NON '(default).

We do the same calculation with ELIM_LAGR =' OUI 'and we check that we get the same result.

The method for eliminating the Lagrange equations is described in [:ref:`R5.03.05 <R5.03.05>`]

The validation is complete:


* We check that the method works for the solvers: MUMPS, LDLT and PETSC


* We check that the method works with non-homogeneous linear relationships


* We check that the solution is well reconstituted on all degrees of freedom, including the Lagrange degrees of freedom (use of the STAT_NON_LINE command).


* We check that the method works on several processors (C and D models)