FIXME

- clay read only the first scop in the chain list

- regenerate access with inequalities

- pip_close in candl (memleak)

- datacopy bug (taken from the polybench symm.c) :

  If we have something like this :
  for (i = 0 ; i < N ; i++)
    for (j = 0 ; j < N ; j++)
      for (k = 0 ; k < j - 1 ; k++)
        a[i][k] = ...

  The generated loop for the datacopy will fail :
  for (i = 0 ; i < N ; i++)
    for (k = 0 ; ; k++) // no end
      a_copy[i][k] = a[i][k]

  This is beacause there is no j in a[i][k] and the matrix domain is copied
  from the original. Clay see that j is not used, and set in the domain :
  j >= 1
  j <= -1
  This will remove the loop j in the datacopy.

  But it's wrong, because k depends of j.