Subject: Re: [boost] [system][filesystem v3] A possible error_code argument compromise
From: Scott McMurray (me22.ca+boost_at_[hidden])
Date: 2009-10-27 10:45:38
2009/10/27 Stefan Strasser <strasser_at_[hidden]>:
>
> boost::optional<error_code&>?
> optional does have implicit conversion and the rest of your public interface.
>
Ah, good. I hoped it did, but I got confused by the Optional
References page in the docs.
>
> optional<T &> is implemented as a container of T & plus a bool though, but I
> see no reason not to create a boost::optional specialization that implements
> optional<T &> by holding a pointer that's exposed as a reference in the
> interface.
>
Which I suppose brings us right back to the question of whether null
references are legal, since that (worthwhile) optimization is only
valid if they're illegal.