Some potential inconsistencies with latest trunk

[vraj004]

I merged my fork with the latest trunk and would like to point out some potential minor issues.

(1) opencmiss.f90 constants for direct solvers are declared but not made public. e.g: CMISS_SOLVER_DIRECT_LU. Is there a reason this was removed?

(2) solver_routines.f90 - Should make sure that cellML models are only evaluated on nodes that have been assigned models.

(3) solver_routines.f90 - CELLML_EQUATIONS_CREATE_FINISH
Used to call SOLVER_CREATE_FINISH at line 800 - any reason why this is removed?

I have made appropriate changes in my fork, but would like to double check if there are reasons behind these observed changes that I am unaware of.

[adamreeve]

Hi Vijay

1. This was never public but probably should be. It looks like you made it public in your commit: vraj004/cm@f8c7deb, but this was never merged into master.

2. Not sure about this, is there some code that has been changed?

3. This wasn't removed either, you added this in vraj004/cm@5e6f8e7, which was never merged into master. I don't think SOLVER_CREATE_FINISH should be called here though, as it should be called from SOLVERS_CREATE_FINISH, which is called from the specific "PROBLEM_SETUP" routine for the problem you're using when PROBLEM_SOLVERS_CREATE_FINISH is called, eg. see BIODOMAIN_EQUATION_PROBLEM_SETUP where PROBLEM_SETUP%SETUP_TYPE is PROBLEM_SETUP_SOLVERS_TYPE and PROBLEM_SETUP%ACTION_TYPE is PROBLEM_SETUP_FINISH_ACTION. All this does now is mark the solver as finished so it can't be changed further, the solver mapping and specific solver type aren't finished until the boundary conditions are finished in SOLVER_EQUATIONS_BOUNDARY_CONDITIONS_CREATE_FINISH.