Bruno Haible <[EMAIL PROTECTED]> writes: > should return "1234321234321234.000000" ... > but returns "1234321234321234.000010". > > Should this be worked around in gnulib, or do you find such unmotivated > rounding errors acceptable?
Personally I'd say it should be worked around, but that it's low priority. What do the relevant standards say? If memory serves, IEEE-754 doesn't require accuracy if you ask for more than 17 digits of output. That might be an argument for leaving it alone.
