What are the hardest cases for likelihood calculation on multivariate gaussian distribution?

I am implementing a function for calculating the likelihood of a point on multivariate gaussian distribution. I want to test it with extreme cases.

function [P,E] = gauss_pdf(X,M,S)
% In:
%   X - Dx1 value or N values as DxN matrix
%   M - Dx1 mean of distibution or N values as DxN matrix.
%   S - DxD covariance matrix
%
% Out:
%   P - Probability of X. 
%   E - Negative logarithm of P
  if size(M,2) == 1
    DX = X-repmat(M,1,size(X,2));  
    E = 0.5*sum(DX.*(S\DX),1);
    d = size(M,1);
    E = E + 0.5 * d * log(2*pi) + 0.5 * log(det(S));
    P = (1 / sqrt(2*pi*det(S))) * exp(-E);
  elseif size(X,2) == 1
    DX = repmat(X,1,size(M,2))-M;  
    E = 0.5*sum(DX.*(S\DX),1);
    d = size(M,1);
    E = E + 0.5 * d * log(2*pi) + 0.5 * log(det(S));
    P = (1 / sqrt(2*pi*det(S))) * exp(-E);
  else
    DX = X-M;  
    E = 0.5*DX'*(S\DX);
    d = size(M,1);
    E = E + 0.5 * d * log(2*pi) + 0.5 * log(det(S));
    P = (1 / sqrt(2*pi*det(S))) * exp(-E);
  end

I do not get errors, however I use this piece of code in extreme cases as well, how can I be sure that it is fine?