Not correctly solving certain quartics about rpolyplusplus HOT 1 OPEN

I've started to notice a similar behavior occasionally for some of my uses as well. This lib is really serves best as a theoretical re-implementation of the Jenkins-Traub Algorithm. The original RPOLY software has dozens of safety catches for numerically unstable corner cases. Specifically, much effort is spent on avoiding division by zeros or near-zeros. I don't have the cycles to implement all the safety checks right now, but if you're interested in debugging it then look for places with divisions like:

https://github.com/sweeneychris/RpolyPlusPlus/blob/master/src/find_polynomial_roots_jenkins_traub.cc#L571
https://github.com/sweeneychris/RpolyPlusPlus/blob/master/src/find_polynomial_roots_jenkins_traub.cc#L637
https://github.com/sweeneychris/RpolyPlusPlus/blob/master/src/find_polynomial_roots_jenkins_traub.cc#L778

Specifically the last link is where I've noticed unstable divisions happening the most.

from rpolyplusplus.