Bugs in Mma functions on the NeXT

*To*: mathgroup at yoda.physics.unc.edu*Subject*: Bugs in Mma functions on the NeXT*From*: herrmann at siam.math.tamu.edu (joe herrmann)*Date*: Wed, 10 Feb 1993 17:07:27 -0600

Dear Mathgroup, I installed Mma 2.1 for the NeXT which I ran under NeXTStep OS 2.1 to fix the bugs in Mma 2.0 that E^(-625000)= 1.40101 10^22, E^(-670000)= 2.51105 10^56, etc. Mma 2.1 fixed these particular values and others I identified but I have found the values given below are still incorrect in Mma 2.1 under NeXTStep OS 2.1. However all exponential bugs I am aware of are fixed in Mma 2.0 or 2.1 by upgrading to NeXTStep OS 3.0. However, I found a bug in BesselJ in 3.0 as can be seen in the following: BesselJ[177.4,28.-.28I]=0. + 0. I and BesselJ[177.4,28.]=0. The first is definitely wrong since all zeros of BesselJ with order greater than -1 are real, moreover for Mma 2.0 on NeXT OS 2.1 and for Mathematica (MIPS) I find -121 BesselJ[177.4,28.]= 2.513042235631 10 -122 -121 BesselJ[177.4,28.-.28I]= -4.56550333659 10 - 2.49427565746 10 I Exponential bugs in Mma 2.1 under NeXTStep OS 2.1: Table[{x,N[E^x]},{x,-300000,-200000,5000}] -130289 {{-300000, 4.5230253774 10 }, -71 {-295000, 4.1807 10 }, -125946 {-290000, 3.9833462329 10 }, -102060 {-235000, 6.2625720561 10 }, -296 {-230000, 1.05042 10 }, -97717 {-225000, 5.5153333723 10 }, Sincerely, Joe Herrmann Texas A&M Math Dept herrmann at siam.tamu.edu