Re: [boost] throw in dtor (Was: Re: smart pointer issues list)

From: Dave Abrahams (abrahams_at_[hidden])
Date: 2000-04-14 19:15:58

on 4/14/00 12:15 PM, Valentin Bonnard at Bonnard.V_at_[hidden] wrote:

> Dave Abrahams wrote:
> 
>> OK, though I don't really think that carries much weight, since it doesn't
>> make any sense to have a throw from a destructor under current language
>> rules.
> 
> Whether it is allowed or not by your personal coding
> conventions, it makes a lot of sens to throw in a dtor

I don't agree with that part. There are some edge cases, but in general it
doesn't make much sense, IMO.

> and never leads to undefined behaviour.

That part is true. The fact that there is no undefined behavior doesn't make
it a reasonable thing to do.

> (It's the same as throwing from a ctor, and carries the

Not quite. A ctor is never implicitly invoked by stack unwinding (it can
only happen if explicitly invoked in a dtor). It carries an attendent risk
of program termination not implied by a throw from a ctor.

> same myths and legends as throwing from a ctor used to
> carry some time ago.)

I don't think so. It's pretty easy to understand why throwing from a dtor is
dangerous. The myths about throwing from a ctor were based on lack of
understanding.

-Dave