On Wednesday 26 June 2002 3:00 pm, Lars Gullik Bjønnes wrote: > Angus Leeming <[EMAIL PROTECTED]> writes: > | On Wednesday 26 June 2002 2:22 pm, Lars Gullik Bjønnes wrote: > >> Angus Leeming <[EMAIL PROTECTED]> writes: > >> | Lars, > >> | > >> | I find that there's no problem in using boost::bind on a class that is > >> | NOT derived from boost::trackable. So what's the point of > >> | boost::trackable? > >> > >> That you can remove/delete the object from the trackable class and > >> have this reflected in the signal (i.e. the slot gets automatically > >> deleted from the signal) > >> > >> same purpose as SigC::Object > | > | I could not connect to a class object that was not derived from > | SigC::Object. Failed at compile time. The fact that I CAN connect to a > | class object that is not derived from boost::trackable suggests that this > | is an optiisation thing only. > | > | Especially since all seems to work fine. > > ... I guess only as long as you do not delete the objects with the > slots...
Then that would be a regression from SigC. I do not see the boost people making such backward steps. Especially since they are sooooooo keen on compile-time checks. Anyway, this is now academic as the relevant classes now derive from boost::trackable. Incidentally, there's no need for the derivation to be public... Angus
