+       if (d > (double)LONG_MAX || d < (double)LONG_MIN)
+               return (-1);

I wonder if there is a libm function which can do the validation / bounding
better than this.  Is there a loss in precision?  Can anything weird happen
when this receives a denorm or other non-complient fp number? 

> the log_warn in ntpd_settime could also be a fatal, after all, but the
> existing ones are also ust warnings.

I think ignoring the condition is enough.

Reply via email to