Hi Mathew,

I've been looking arround to the shared_ptr<> and my impression is that is
not thread safe.
If I'm righ, are you thinking of adding the possability of making it thread
safe through a policy or to provide another implementation?


Regards
Cláudio Albuquerque

"Cláudio Albuquerque" <cláudio@nowhere.com> wrote in message news:fcqu9l$27uq$1@digitalmars.com...
>
> Hi Mathew,
>
> I've been looking arround to the shared_ptr<> and my impression is that is not thread safe.

That's correct.

Although I'd probably phrase it that its thread-safety is not defined. ;-)

> If I'm righ, are you thinking of adding the possability of making it
thread
> safe through a policy or to provide another implementation?

I wasn't, but I'm happy to consider it.

I'm hoping to get some STLSoft-time this w/e. If you want to make any suggestions, we can maybe get them in the next release ...

Cheers

Matt

Hi Mathew,

Well I due like policy allot, maybe too much, the problem there is that I end up with too many template parameter's ... but that's another story.

Mentioning another lib that I use (but ours is still my favorite one ;-) ) I do like the idea behind Loki's pointers but in the end It's over to "complicated", although due like the power and flexibility. (This reminds me of the first pages of a nice book I'm reading now, Extending STL, that say exactly that, if it's too complicated to use or to discover people tend to stay away or build their own...)

So in this case my suggestion would go to a template parameter that would state the policy for multithread or not.

PS: By the way did you forget the scoped_method<> ... ?

Regards
Cláudio Albuquerque


"Matthew Wilson" <matthew@hat.stlsoft.dot.org> wrote in message news:fcui6p$811$1@digitalmars.com...
>
> "Cláudio Albuquerque" <cláudio@nowhere.com> wrote in message news:fcqu9l$27uq$1@digitalmars.com...
>>
>> Hi Mathew,
>>
>> I've been looking arround to the shared_ptr<> and my impression is that
>> is
>> not thread safe.
>
> That's correct.
>
> Although I'd probably phrase it that its thread-safety is not defined. ;-)
>
>> If I'm righ, are you thinking of adding the possability of making it
> thread
>> safe through a policy or to provide another implementation?
>
> I wasn't, but I'm happy to consider it.
>
> I'm hoping to get some STLSoft-time this w/e. If you want to make any suggestions, we can maybe get them in the next release ...
>
> Cheers
>
> Matt
>
>