Alfredo Braunstein wrote:
> Angus Leeming wrote:
>
>> I don't see the bug. You return to the start of the list, I move to the
>> next itme in the list. The 'prev' stuff is safe if ugly, as I understand
>> list.
>
> Sorry, I was away.
> What if for instance, *prev has dissapeared in the time being? (because of
> multiple calls to kill() in the function called by emiting the signal).
> Are you making some assumption on the implementation of list? The list
> could even become void.
> I think that you cannot make any assumption on the state of the list,
> after emiting the signal.
>
> Regards, Alfredo
Ooohhhh, that's devious. In that case I think we really need a two pass
algorithm:
ListType::iterator it = forkedCalls.begin();
ListType::iterator end = forkedCalls.end();
for (; it != end; ++it) {
bool remove_it = ...;
if (remove_it) {
// Emit signal.
// Do not remove the item from the list yet.
(*it)->emitSignal();
delete *it;
*it = 0;
}
}
// Check whether any of the list contents have been deleted.
ListType::iterator it = forkedCalls.begin();
while (it != end) {
if (*it) {
++it;
else
it = forkedCalls.erase(it);
}
Thoughts?
--
Angus
