[PATCH] Errors with bern() and zerobern:false



Unrelated observation about code in bffac.mac: The function cbffac is inaccurate in portions of the left half plane; for example

  (%i11) cbffac(-2.0b0 + 1.0b0*%i,25);
  (%o11) 3.264827482100833639193234b-1*%i-1.715329199082726787943681b-1

In this case, the function gamma does a reflection before calling cbffac--I think (%o12) is accurate

  (%i12) gamma(-2.0b0 + 1.0b0*%i);
  (%o12) 1.339097176053258b-1-9.628651530237881b-2*%i

The user documentation for cbffac doesn't mention anything about accuracy:

  (%i13) ?? cbffac;
   -- Function: cbffac (<z>, <fpprec>)
       Complex bigfloat factorial.
      `load ("bffac")' loads this function.
 (%o13) true

--Barton