Re: [boost] shared_ptr implementation without template friends

From: Greg Colvin (greg_at_[hidden])
Date: 2000-07-31 19:30:24

Sorry, but in my opinion Boost has already gone above and beyond
the call of duty to make things work with broken compilers.  I
don't see why we should also bend over backwards to protect
incompetent programmers.  And yes, I am in a foul mood today ;->

----- Original Message ----- 
From: Neil Stewart <nwstewart_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Monday, July 31, 2000 3:46 PM
Subject: Re: [boost] shared_ptr implementation without template friends

> ----- Original Message -----
> From: David Abrahams
> 
> > This sounds to me like the way the MSVC feature works. Doesn't the
> > auto-completion find private members, too? I can't seem to get that
> feature
> > to work for me in VC right now, so I can't test it, but I wouldn't want to
> > be compelled to mangle all of my private names so that people wouldn't use
> > them.
> 
> It does complete private members, as do many editors (quite useful when
> you're writing the class's members). The point is, though, that the compiler
> will generate an error if they do access the member illegally. What I have
> been trying to suggest is a way to make this happen for public members that
> should really be private (but cannot be on broken compilers). I wouldn't
> dream of suggesting doing this with any private members.
> 
> > Give those people a slap on the wrist. Even if they randomly find a member
> > name, how will they know what it means without reading the source? If they
> > look at the source it will be revealed as a member that "should be
> private".
> 
> Hehe, some people can take quite a slapping before the message sinks in.
> 
> I agree fully with the sentiments people have expressed here. A warning
> comment *should* be sufficient, but people do take shortcuts (especially
> with deadlines looming), and in many cases with the best of intentions.
> 
> The problem I have with what people are saying here is that they are not
> criticising the method I was suggesting (on the whole), but the notion that
> it needs to be done at all. I'm not that fond of the method either, but I
> was hoping to see what people thought could be done in that direction to get
> a similar result, if possible.
> 
> - Neil.
> 
> 
> 
> 
> 
> 
>