On Mon, Sep 12, 2016 at 02:02:33PM +0200, Ross Finlayson wrote: > Rather than making an obscure reference to a 6-year-old message (that was > likely in response to something completely different to what you’re concerned > about), why don’t you just ask the question you really want to ask? That > way, we’ll all know what your question really is...
To be honest, I was debugging code that uses exactly that advise from that 6-year-old message to limit the number of active sessions (but I only discovered that message way later in my debugging). It took me a while figuring out that returning NULL from createNewClientSession causes createNewClientSessionWithId to insert a NULL session into the HashTable fClientSessions. This wouldn't be too bad, but when you destruct a GenericMediaServer, it turns out that fClientSessions is only deleted until finding the first NULL pointer. Thus actual ClientSessions can be leaked. While that also looks like a simple resource leak, ClientSessions have a fLivenessCheckTask that eventually fires. When it does, a use-after-free (on the now deleted GenericMediaServer) can cause very weired behaviour. That is what lead me to the initial question: Is it actually ok for createNewClientSession and others to return NULL? If it is, then you have a bug. If it isn't, I have a bug (and that 2009 message is no longer accurate). So whose bug is this? Helmut _______________________________________________ live-devel mailing list live-devel@lists.live555.com http://lists.live555.com/mailman/listinfo/live-devel
