From: Daniel Frey (d.frey_at_[hidden])
Date: 2008-04-26 11:58:31
On Sat, 2008-04-26 at 18:34 +0300, Peter Dimov wrote:
> Daniel Frey:
> > When looking at the test case that fails with .use_count() != 0, I don't
> > really see why the call to shared_from_this() in line 130 should throw.
>
> The test looks correct to me. After early_px.reset() and px.reset(), no
> shared_ptr owns &x. Their initial use_count() is 2, so after the two resets
> a weak_ptr to &x should expire (it might be a good idea to add it to the
> test). We'll indeed need weak_count::empty to make this and test5 both pass,
> though.
The current test protects against misuse of shared_from_this, the
exception it expects is just a diagnostic for it. My new version even
allows one more use case:
When the refactored implementation reaches line 130, the
weak_count.use_count() == 0 is true, thus init_weak_once() initializes a
fresh _shared_count/_weak_count pair and expects someone else to take
over ownership later as if the object was just created. Unless I'm
missing something...
Regards, Daniel