Submitter: Fred J. Tydeman
Submission Date: 2015-01-07
Document: WG14 N1903
"A range error occurs if x is too large." is misleading (or ambiguous) for expm1 (220.127.116.11p2), erfc (18.104.22.168p2), and lgamma (22.214.171.124p2).
"too large" could mean either +/-large value (in which case "too small" means +/-near zero) or just +large value (in which case "too small" means -large value).
126.96.36.199p2: expm1(-DBL_MAX) is -1, which is not a range error.
188.8.131.52p2: erfc(-DBL_MAX) is 2, which is not a range error.
184.108.40.206p2: lgamma(-DBL_MAX) is a pole error, which is not a range error.
Suggested Technical Corrigendum
Add the word "positive" before x in those three cases so that they are:
A range error occurs if positive x is too large.