Hi Ken,

> > It determines the style of date(1) to invoke without relying on user
> > input, then when that's known it invokes it again with the desired
> > number of seconds but without discarding stderr in case of problems,
> > and the existing `set -e' would exit on an error.
> It occurs to me that feature tests are exactly what autoconf has been
> designed to do, so maybe we should put the smarts in there?

Sounds right.

Cheers, Ralph.


Reply via email to