[boost] Re: Re: Re: Re: Re: Threads & Exceptions

From: Eric Woodruff (Eric.Woodruff_at_[hidden])
Date: 2002-08-13 18:55:12

Do you not see the shared_ptr policies as being useful, or even necessary
sometimes? The templated copy constructor is an attempt at faking
inheritance.

Personally, I think it would fix all of the exception issues. Many things
revolve around shared_ptr in my opinion (I also think that there should be a
shared_ptr boost::New<> as I've probably mentioned before.)

I hate to give away my evil scheme but...

My ideal usage would be to

throw boost::New<std::invalid_argument> ("opinions are subjective");

from a generic function that could be in a thread or anywhere else. (Well I
wouldn't use it like that, I would have an assertion template and sanity
checking probably, but that would be the effective implementation inside the
templated assertion.)

Of course this introduces all sorts of other problems, but they can possibly
be ironed out ;)

If anyone can make this work, I will gladly shake their hand.

----- Original Message -----
From: David Bergman
Newsgroups: gmane.comp.lib.boost.devel
Sent: Tuesday, 2002:August:13 7:28 PM
Subject: RE: Re: Re: Re: Re: Threads & Exceptions

Does it not seem a bit overkill to replicate the whole C++ at a meta
level :-?

The only thing I wanted out of this discussion was to have the thread
terminate as the default behavior, instead of the whole program. That
now seems to be the asymptotic consensus in this meta-thread, so I am
satisfied.

If I care that much about the exception, I would probably have
implemented a more direct thread communication mechanism (e.g., shared
memory + monitors), which would, obviously, require me to be the
implementer of that "other thread".

The case is that we cannot know the exact reason for an independent
thread's death, or exceptional behavior, since the "language of the
thread is foreign". Even if we could overcome this situation by some
clever extra-language tricks, why would we? If we managed to have the
bits copied properly in our thread space, what sense would it make?

I kind of liked the idea of having the "what" copied somewhere (was it
Bill?), for logging and analytical purposes. That text is probably the
best info we have in the general case.

/David

-----Original Message-----
From: boost-bounces_at_[hidden]
[mailto:boost-bounces_at_[hidden]] On Behalf Of Eric Woodruff
Sent: Tuesday, August 13, 2002 7:06 PM
To: boost_at_[hidden]
Subject: [boost] Re: Re: Re: Re: Threads & Exceptions

What about an inheritance policy that the user could change?
inheritance, no
inheritance, multiple inheritance...

----- Original Message -----
From: David Abrahams
Newsgroups: gmane.comp.lib.boost.devel
Sent: Tuesday, 2002:August:13 6:25 PM
Subject: Re: Re: Re: Re: Threads & Exceptions

From: "Eric Woodruff" <Eric.Woodruff_at_[hidden]>

> Would it be possible to implement the shared_ptr so that it could
inherit
> from a shared_ptr of a second parameter  (yes, dual hierarchies, but
no
> multiple-inheritance I guess), so that it didn't have a number of
reference
> counters dependent on the height of the hierarchy?

No

_______________________________________________
Unsubscribe & other changes:
http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________
Unsubscribe & other changes:
http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________
Unsubscribe & other changes:
http://lists.boost.org/mailman/listinfo.cgi/boost