Peter Memishian wrote:
>  > Changing when daemonization took place was a suggestion to fix the race.
> 
> Right, and I thought your concern was that daemonizing later would make it
> possible that there were pending door calls across the fork() that we'd
> have to deal with.  But by definition (since the service hasn't come
> online yet), those door calls must be spurious.

My understanding (Dan, correct me if I'm wrong) was that Dan's concern 
was with having two processes simultaneously servicing the same door 
call, which might cause damage to the system regardless of whether those 
calls were spurious and end up failing.

-Seb

Reply via email to