[net.lang] f77 bug

jabarby@wateng.UUCP (Jim Barby) (09/15/83)

It appears f77 that if you initialize a double precision (or real)
using a data statement to a number greater than the largest number
f77 crowbars it to the largest number WITHOUT telling you (ie.
without a single warning or error message).

For example

	double precision xxx
	data xxx/1d50/

will result in xxx = 1.701411834607d38 NOT 1d50 as desired.

dmmartindale@watmath.UUCP (Dave Martindale) (09/16/83)

For what it's worth, the silent replacement of a too-large number with the
largest possible floating value is a "feature" of ldexp(), which is what
the Berkeley version of atof() calls to do the final conversion of the
integer mantissa and exponent to floating format.  It DOES set "errno"
in the process, but no one checks it.