I also noticed several tests will fail and can even isolated the data to run a reproducible failure, but I think those are just rounding errors. Perhaps it should relax the tolerance.
On Sat, Jun 30, 2018, 9:59 AM Bill Heagy <[email protected]> wrote: > If I run test/g410.ijs often enough, I get: > > 100 RUN2 'g410' > ....... > 1 1 1 1 1 1 1 1 > (g = gamma) x=: _4.1 _4.9 4.1 4.9 j./_0.4 _0.5 0.4 0.5 _4 _5 4 5 > 1 1 1 1 1 1 1 1 > 1 1 1 1 1 1 1 1 > 1 1 1 1 1 1 1 1 > 1 1 1 1 1 1 1 1 > > (g =!.1e_12 gamma) 1.5 > 1 > (g =!.1e_12 gamma) 1r3 1r2 2r3 > 1 1 1 > (g =!.1e_12 gamma) x=: 1e_2*_7e2+?2 20$2e3 > 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 > 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 > (g =!.1e_12 gamma) x=: 3e_2*_7e2+?2 20$2e3 > 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 > 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 > (g =!.1e_12 gamma) z=:j./1e_2*_7e2+?2 20$2e3 > 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 > |assertion failure > |[-103] /home/wheagy/git/jsource/test/g410.ijs > > Which specific value is wrong changes from failure to failure, but it's > always this line, and it appears to happen about every 50 to 100 > iterations. This happens on debian linux 32 and 64nonavx, with the > latest beta release and when compiled from the latest version on github. > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
