Re: [boost] [smart_ptr] questions about some regression test cases

From: Frank Mori Hess (fmhess_at_[hidden])
Date: 2008-04-27 20:27:38

On Sunday 27 April 2008 15:02, Daniel Frey wrote:
> My questions result from the observation that in the virtual dtor of
> enable_shared_from_this, I tried BOOST_ASSERT( _shared_count.empty() ),

Are you aware that will abort the program in the case of a derived 
constructor calling shared_from_this() then throwing?  It might be worth 
considering making it a requirement that no constructor can throw after 
calling shared_from_this though, see the first half of an earlier post:

http://lists.boost.org/Archives/boost/2008/04/135372.php

> but this makes 3 tests fail. Looking at these tests, I don't see why
> they should be legal:
>
> In esft_constructor_test.cpp, line 143, X initializes early_px by
> calling shared_from_this, but X is never passed to a shared_ptr which
> then owns it. From what I understand, this should be illegal.
>
> The next problem is esft_constructor_test.cpp, line 160.
> y.shared_from_this() returns a shared_ptr, but again, no shared_ptr
> takes ownership of Y later. While shared_from_this() is not called from
> the ctor, I wonder if this is yet another use case or just an accident.
>
> The test case in shared_from_this_test.cpp, line 135 is similar.

-- 
Frank