Re: [sigc] Unexpected behavior when freeing a trackable object from inside a slot

[Kjell Ahlstedt <kjellahlstedt gmail com>]

Den 2019-08-14 kl. 17:09, skrev Andrey H. via libsigc-list:
> Hi,
>
>     I noticed that if a trackable object is freed from inside the slot 
> which is setup to track this object then the slot is immediately 
> invalidated. Problem arises when such a slot holds a lambda which 
> captures some other variables by-value or some variables are bound by 
> value to the slot. In such cases the invalidation of the slot causes 
> all bound object instances to be immediately freed.
>
>     I have attached a test case patch (for the current master) which 
> demonstrates the issue using lambda captures and sigc::bind bounds.
>
>     My question: is it a bug or it's supposed to work like that?
>
>     The patch also contains a 3rd test which disconnects the slot 
> explicitly from inside the lambda slot with captured variables, but in 
> this case it doesn't cause captured variables to be freed right from 
> inside the slot. So I'm inclined to think that track_obj should behave 
> in a similar safe way.
>
> BR, Andrey
A bug or not a bug? It's not ideal, of course. But I think it would be 
difficult to make sigc::track_obj() postpone invalidation the way 
slot::disconnect() does. It postpones some actions until the slot call 
has finished. Disconnecting the slot is less rude than deleting part of 
it while it's executing.

You can file an issue to GitHub if you like, but I think this is 
something you have to accept. Unless you can create a clever fix yourself.

Kjell