cg_solve error

[duncan]

Jane got this error. She's working on a simulation where it seems that least squares gives incorrect fitted values. [There isn't enough info here to reproduce the bug sorry.]

Traceback (most recent call last):

  File "Z:\inundation\sandpits\jsexton\ANUGA\production\sydney_2006\run_sydney_smf_polyingo.py", line 223, in ?

    ,verbose = False

  File "z:\inundation\sandpits\jsexton\ANUGA\inundation\caching\caching.py", line 349, in cache

    T = apply(func,args,kwargs)

  File "z:\inundation\sandpits\jsexton\ANUGA\inundation\pyvolution\least_squares.py", line 124, in fit_to_mesh_file

    precrop = precrop)

  File "z:\inundation\sandpits\jsexton\ANUGA\inundation\pyvolution\least_squares.py", line 215, in fit_to_mesh

    vertex_attributes = interp.fit_points(point_attributes, verbose = verbose)

  File "z:\inundation\sandpits\jsexton\ANUGA\inundation\pyvolution\least_squares.py", line 942, in fit_points

    f[:,i] = self.fit(z[:,i])

  File "z:\inundation\sandpits\jsexton\ANUGA\inundation\pyvolution\least_squares.py", line 916, in fit

    return conjugate_gradient(self.B, Atz, Atz, imax=2*len(Atz) )

  File "z:\inundation\sandpits\jsexton\ANUGA\inundation\utilities\cg_solve.py", line 65, in conjugate_gradient

    d = r + bt*d

OverflowError: math range error