Issue #9599:  Further accuracy tweaks to loghelper.  For an integer n that's small enough to be converted to a float without overflow, log(n) is now computed as log(float(n)), and similarly for log10.
2 files changed