Alan, Could this be some sort of a compiler optimization that may be causing this? May be some memory barrier is required? Padam Alan DeKok wrote:
fabiana marvani wrote:
Why is not there a protection for "null pointer" to avoid this kind of problem?
Because the design of the server means that this crash *should* be impossible.
The request packet has been placed in a hash table. The crash comes because the request data structure is still in the hash, but the request packet has been freed. This should *not* happen.
And reading src/main/event.c, there are only 3-4 locations in the code that reads/writes "request->packet". i.e. it FIRST removes the request from the hash, and THEN frees the packet. It's always in this order, and always done in the main thread.
So I have no idea why this is happening...
Alan DeKok. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html