On 2020-11-09 09:39:25 +0000, Geoff Clare via austin-group-l at The Open Group wrote: > I believe this is a misuse of "negative". The RETURN VALUE section > says "For finite values of x < 0, and finite non-integer values of y, > a domain error shall occur ..." > > So the ERRORS section should say "less than 0", not "negative". > (It should also say "a finite value" to match RETURN VALUE.) > > Likewise, the two Pole Error descriptions in the ERRORS section should > also say "less than 0" instead of "negative".
The word "negative" should normally be equivalent to "less than 0". Interpreting it as "of negative sign" is unusual, and apparently not defined by POSIX. So, in particular, there is an issue with the drand48() and erand48() definition, which is ambiguous when a zero is returned, given the fact that +0.0 and -0.0 are equal. For log, etc., it is said: Domain Error The finite value of x is negative, or x is −Inf. and this is supposed to exclude -0. -- Vincent Lefèvre <vinc...@vinc17.net> - Web: <https://www.vinc17.net/> 100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/> Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
