Please ignore this whole thing. Sheesh, been too long since I've done low level sockets stuff. The end problem was a bit of sloppy code and me misremembering an issue with Win32 specific servers that doesn't apply to linux/osx.
> Hi Folks, > > Maybe someone has done this before so figured I'd ask while I'm > still experimenting with possible solutions. In the process of finishing > up > some porting to OsX I noticed that I have a race condition in my code > related to listener sockets. Previously I would just bring up the > listener > socket, bind it, make it non-blocking, call accept once (dropping any > success which should never happen anyway) and then query the port before > moving on. > > Now, with libevent I'm curious if there is a better method of doing > this. I started out just using the evconnlistener_new_bind but this is > obviously not calling accept immediately on the socket since OsX is > functional only if I insert a couple calls to libevent_loop into the code > prior to requesting the bound port. (NOTE: Win32 works fine but I assume > that is because it calls AcceptEx immediately on the socket.) In a single > threaded system this is not a problem but much of this portion of my code > is > actually running in a worker thread owned by a multi-cast dns (i.e. > bonjour/avahi) wrapper. > > Currently my best guess is to go back to the old method and just use > evconnlistener_new instead of the '_bind' variation passing in the socket > after I've gotten the required information out of it. > > Any suggestions are welcome. Preferably in the future, listener > might be able to do this fully asynchronously with an event callback or > something and then it would all work without any extra front end work or > portability issues. > *********************************************************************** To unsubscribe, send an e-mail to majord...@freehaven.net with unsubscribe libevent-users in the body.
