DYAEB@SLACVM.SLAC.STANFORD.EDU (David Aston) (03/15/91)
The definition of DBL_MAX in <float.h> seems to be in error. It is very slightly different from the value MAXDOUBLE in <math.h> such that it maps to Infinity instead of the max representable normalised double. <float.h> is the same in versions 1.0a and 2.0 of the system software. But wait, there's more! In Release 2.0, DBL_MAX (and apparently any number larger) is mapped to -Infinity _not_ +Infinity. This is a "feature" of Release 2.0. I have tried an 030 with 1.0, an 030 with 2.0 and an 040 with 2.0. Programs built with Release 2.0 have the sign reversal, but do not when built with 1.0 (even when run under 2.0). I also believe that FLT_MIN is not behaving correctly, I'll post again when I've sorted it out. I have little C/Unix experience, so would appreciate feedback about this "feature". -- This opinion is free. You get what you pay for. BitNet: dyaeb@slacvm I do not speak for SLAC, Stanford U, DOE or Dan Quayle.