Re: mutex_traits<>

From: bill_kempf (williamkempf_at_[hidden])
Date: 2002-01-31 15:10:09

--- In boost_at_y..., Iain.Hanson_at_u... wrote:
> 
> Author:  david.abrahams (david.abrahams_at_r...) at unix,mime
> Date:    31/01/02 19:09
> 
>      
> >...so you give newbies a "safe_mutex" on which scoped_lock works 
and for 
> >which there are no lock/unlock functions? Maybe I'm just getting 
hung up on 
> >the principle of the thing. After all, even experts are likely to 
use 
> >scoped_lock most often.
> 
> One of the things that most concerns me with the totally safe 
interface is 
> that if I want to unlock the mutex early and then re-aquire it 
later I 
> can't.

This isn't true.  The Scoped Lock classes provide the ability to 
unlock and then later lock the mutex.  This covers the _vast_ 
majority of uses (and probably _all_ of the issues that the 
proverbial "newbie" is going to need).  The only limitation is the 
scoped nature of the lock, which prevents efficient lock patterns for 
things such as a lock_ptr<>.  That's what the "traits" approach is 
supposed to address, though.

> I have to create two mutex's :-( with all the extra complication 
that 
> entails in other functions .

I can not envision any case in which this would be the case (even 
with out the traits).  Can you post an example?

Bill Kempf