Bug#798601: Bug Analyzation

[Roelof Berg]

I assume, the software is ok but the test automation is not. The 
'algorithm under test' gives a bit fuzzy results depending on the 
multicore calling order and depending on numerical rounding issues. 
Therefore the automated tests accept results within a range of plausible 
values.  Seems that one of the automated tests is too strict. This is 
probably best fixed by upstream (which is also me, have I told about my 
multiple personality disorder yet ? :).

Regards, Roelof