problems with ~G in format

[Nikodemus Siivola]

Originally reported as: http://article.gmane.org/gmane.lisp.steel-bank.devel/13878

CL-USER> (format nil "~G" most-positive-double-float) ; totally messed up
"179769313486231570000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000. "
CL-USER> (format nil "~G" 0.04165461138601871d0) ; extra zero
"4.16546113860187140d-2"
CL-USER> (format nil "~G" 9.6000000000000474d-2) ; two extra zeros
"9.60000000000004700d-2"

[Douglas Katzman (dougk)]

I think most-positive-double-float is NOT totally messed up.
Trying to follow the algorithm verbatim in
http://www.lispworks.com/documentation/HyperSpec/Body/22_ccc.htm
I calculate the ~F parameters that would produce that long string.
The only other option is that in ~F it says "If w is omitted, then if the magnitude of arg is so large ... that more than 100 digits would have to be printed, then an implementation is free, at its discretion, to print the number using exponential notation instead"

At any rate this hardly qualifies as medium priority after 9 years of idleness.