From: Peter Dimov (pdimov_at_[hidden])
Date: 2001-11-12 15:24:50
From: "Stewart, Robert" <stewart_at_[hidden]>
> Why not use a set_terminate_handler-like interface for setting the
> boost_error-like function? If that works for the standard, why not for
> Boost? I suppose that you're worried that the function pointer might not
> initialized to the address of boost_error when called from some static
> object's initialization. If you initialize the pointer to zero, and use
> that as a special case that calls boost_error, then there should be no
> problem (other than that the library user may have wanted to have their
> replacement called at a point when their "set handler" logic hasn't run.
Yes, this is exactly the problem that is solved by the current design.
-- Peter Dimov Multi Media Ltd.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk