Things to do for this thorn
===========================
$Header$

Jonathan Thornburg:
* make the hard-wired 1.0e-6 finite differencing grid spacing in
   Bona_Masso_data.F77  a parameter
* Are we setting the initial data at the right time level
  (there are comments in  gauge.F77  that we're half-a-time-step off)?
  Is this still a problem now that we always use MoL?
  Do we still need to support the non-MoL case?
[Note by Erik Schnetter, 2005-05-15: This is not an issue any more.
Jonathan, please remove this item.]
* add options to set ALL time levels of initial data, not just the
  current level

----------------------------------------

Ian Hawke:
It would be nice to be able to output the difference between a
numerical evolution and the exact solution.

----------------------------------------

Sascha Husa:
It would be nice if Exact could set up variables other than the ADM ones
(indeed it appears to not even grok BSSN right now).

[Note by Erik Schnetter, 2005-05-15: This is a misunderstanding by
Sascha.  Thorn Exact works just fine with the BSSN variables.
Jonathan, please remove this item.]

----------------------------------------

Sascha Husa:
The boundary setup needs to be generalized to handle ghost zones wider
than 1 point.

[Note by Erik Schnetter, 2005-05-15: This is a misunderstanding by
Sascha.  Thorn Exact works just fine with more than one ghost zones or
more than one boundary point.  Jonathan, please remove this item.]

----------------------------------------

Every model should have .ge. 1 test case!  Right now most of the
models have no test cases at all, and this makes it way too easy for
bugs to creep in and go undetected for a long time.

For models satisfying the Einstein eqns, it would be nice to also have
test cases which compute the Hamiltonian constraint (and maybe the
momentum constraints too) and check that it (they) are suitably small.
	(The down side of that is that is that those test
	cases would then need at least ADMConstraints
	to be compiled into your Cactus binary, and that
	in turn requires ADMMacros.)
This would help guard against typos, wrong formulas, and other
"consistent, but wrong" cases.

----------------------------------------
It would be nice if Exact supported fisheye.
Failing that, it could at least check if FishEye is active
  and cleanly report an error.
----------------------------------------
From schnetter@aei.mpg.de Tue Jan 18 19:24:05 2005
Date: Tue, 18 Jan 2005 19:23:06 +0100
From: Erik Schnetter <schnetter@aei.mpg.de>
To: numrel@aei.mpg.de
Subject: [Numrel] Uninitialised variables AEIThorns/Exact

I noticed that not all metric routines set the variable "psi", which is 
probably some kind of conformal factors.  However, some do, and I 
cannot see a rule for when this is the case.

In Bona_Masso_data.F77, there are later some unconditional calculations 
on this value of psi.

How is this variable psi supposed to be handled?  Should it be 
initialised to some default?  Is there some kind of per-metric flag 
that determines whether it should be set?

There is also a strange statement

      if (psi .gt. 0.0d0) then
        psi_on=1
      else
        psi_on=0
      end if

near the beginning of Bona_Masso_data.F77 which I do not understand.

-erik

-- 
Erik Schnetter <schnetter@aei.mpg.de>   http://www.aei.mpg.de/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.

    [ Part 2, Application/PGP-SIGNATURE  196bytes. ]
    [ Unable to print this part. ]
----------------------------------------