Paul sent me the following issue with the use of copysign in Christians alternative method which is described on this page. Since I don't have much expertise with C++ code and haven't used copysign myself I will pass on Pauls comments:
The article mentions an alternative method using _copysign(x,y) but a common behavior of this function is to return (+1) when y=0 given the following matrix
1 0 0
0 1 0
0 0 -1
this alternate method produces different results
Q.x = sqrt( max( 0, 1 + m00 - m11 - m22 ) ) / 2;
Q.x = _copysign( Q.x, m21 - m12 )
producing Q.x = 0.7071068 whereas the original algorithm:
qx = (m21 - m12) / ( 4 *qw )
here it is clear that when (m21 - m12) equals (0) then (qx) will be zero if you use a version of copysign that returns +1,0,-1 then it might work.