1) 2) How does the lock protect framedrop_policy? And what does it protect? From what I can see there's no need to hold the lock. The implementation of the policy has its own internal synchronization.
3) I could not come up with a reasonable test against deadlock that doesn't involve modifying production code
@Chris, yes definitely a band-aid for now. I think applying a similar custom lock as in Alan's deadlock free branch will be needed for alarm callbacks.
@Daniel
1) 2) How does the lock protect framedrop_policy? And what does it protect? From what I can see there's no need to hold the lock. The implementation of the policy has its own internal synchronization.
3) I could not come up with a reasonable test against deadlock that doesn't involve modifying production code
@Chris, yes definitely a band-aid for now. I think applying a similar custom lock as in Alan's deadlock free branch will be needed for alarm callbacks.