Submitter:Fred Tydeman
Submission Date: 2018-09-25
Document: WG14 N2302

Summary

When long double is a pair of doubles, what is nextafterl(1.L,2.L)? The same issue applies to nexttowardl(1.L,2.L).

Several possible values are:

• 1.L + LDBL_EPSILON
• 1.L + LDBL_MIN
• 1.L + LDBL_TRUE_MIN
• Something else

Suggested Technical Corrigendum:

Add the following to paragraph 2 in clause 7.12.11.3 The nextafter functions

For finite results, the returned value shall be either a zero, a subnormal floating-point number, or a normalized floating-point number.