I can reproduce hangs (but not crashes) for these in the latest version.
There have been some interesting recent developments in polynomial solving, which I planned to implement into je, but they only work for reals, so no help here.
However, I wonder about the possibility of using solely bisection with complex intervals. As is usually the case, it's likely to be less pretty, but faster and more robust than the analytic solution we currently use.
On Sat, 4 Feb 2023, Schmidt-Gröttrup, Markus wrote:
A numerical error crashes J p. _1E10 _1 0 0 1 works p. _77 _1 0 0 0 1 works p. _78 _1 0 0 0 1 crashes p. _14 _1 0 0 0 0 1 works p. _15 _1 0 0 0 0 1 crashes p. _8 _1 0 0 0 0 0 1 crashes p. _7 _1 0 0 0 0 0 1 works Version used: J807/j64/windows Greetings, Markus ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
