Re: Fixed: Segfaults in 3.14 branch since merging ca-over-tcp

[Ralph Lange <Ralph.Lange@bessy.de>]

 On 17.08.2010 21:00, Jeff Hill wrote:
> Looks like a potential issue.
>
> option 1)
>
> You might insert "this->chanTable.removeAll ()" after line 295 in ~cac
> which will cause cac::transferChanToVirtCircuit to noop return at line
> 583. You would need to supply a temporary (auto) tsSLList<T>  target to
> removeAll but could safely ignore
> [...]
>
> option 2)
>
> Just add a bool flag to cac as you propose, set it just after line 295
> in ~cac, and check it at the beginning of cac::transferChanToVirtCircuit.
> It would also need to be initialized in the cac ctor. Of course the set
> and test will need to be protected by a Guard. The downside is that
> this requires slightly more memory at runtime, but this is no-doubt
> insignificant. Perhaps the only downside is needing to remember the
> purpose of yet another binary flag a few years down the road.

I'll go with the simple approach.
This is a one-way communication, so a proper set-and-test is not necessary.
I will set the flag (under the lock) at the beginning of the cac dtor 
and check it (under the lock) at the beginning of 
cac::transferChanToVirtCircuit - that should do it.
I think a comment will reliably remove the need to remember the purpose 
of the flag. ;-)

Thanks for your help!
Ralph