Error while executing service

Hi @benjaml.

What is the actual error that you face? I only see a callstack to where it happens, but no information about what happens.

342 is indeed very high.
The reason for the warning once the queue goes beyond 100 elements is that even that is already an extraordinary high number.

With the exception of the caller having passed a huge (100kb or more) message to opRaiseEvent() right before the warning (in which case the warning can be ignored) having so many queued commands is a very strong indicator that there is a problem in your code. opRaiseEvent() adds commands to the outgoing queue and service() grabs the commands in the queue and sends them out. Messages that would results in more than 1200 bytes of UDP datagram size get fragmented into multiple commands. Hence if a message is really big, then a lot of commands will be added to the queue. However normal messages will just result in 1 or at max a handful of commands. Hence if the queue grows to 100 commands nonetheless, this usually means that your code has not called service() for quite a while / did a lot of opRaiseEvent() calls since the last service() call.

The only ways I see that the queue could be bigger after a service() call than it was before that call, is that you call opRaiseEvent() from within a callback (which is fine) or from a different thread than the one on which you call service() (don't do this without applying locks! - the Photon Client is not thread safe in the sense that you could access the same instance of a class from multiple threads in parallel - it is only thread safe in the sense that static resources which might be accessed by different instances from different threads in parallel, like the memory management, are thread safe, so that you can access different instances from different threads in parallel).


That means you enqueue 340 commands within 2-3 service() calls.
Either you call service() way too rarely (you should call it regularly, multiple times per second - usually 10-20 times a second is a good choice) (please verify that the thread which calls service() is not blocked by i.E. a loading screen for multiple seconds) or you simply send way too much data. Photon is not a file-server, don't use it as such.

Hi @benjaml.


Because by doing so it might cause a buffer overrun in the outgoing buffer of the socket.

I have just checked the code: Each service() call will only pass one UDP datagram to the socket and any commands that don't fit into it that datagram will have to wait until the next call.
You can call service() multiple times in one frame, even in a loop until all commands are out, after you have just passed a huge message to opRaiseEvent(). Each call will then check again if the socket has already sent out the data from the previous call so that we can pass it more data. However you need to add a sleep() call after each service() call to give the socket a chance to actually send out the data between service() calls.

Still 350kb is way beyond what Photon is intended to be used for. Splitting it up into multiple chunks yourself is probably a good idea when you use UDP and try to send more than 100kb. However if you split the data up yourself anyway, it might make sense to split it into something like 1150 byte or so per chunk so that Photon does not have to fragments the chunks. Once you have multiple chunks your code has to deal with the situation that one chunk might get lost and needs to be repeated, anyway.

All this still does not explain the memory access violation within EnetCommand::operator=(). I don't see why this would happen and I am afraid that I would need you to provide reproduction code + information on which platform this is happening to be able to look into this issue and find out what is causing it.