Hi people! Oskar and I have fixed a lot of bugs. We've overhauled the request handling states, and improved the connection and thread limiting (you shouldn't get refused local FCP connections anymore). There have been some fixes to the datastore making it more reliable. We have corrected some bugs in the routing table that could make the node fail to route a request. The bottom line is that request handling should be significantly less flaky now -- and if it is flaky, it is because of easily fixable mistakes in this new code, rather than architectural flaws. Finally, we have implemented a mechanism whereby the node will refuse to accept incoming requests from nodes running older versions of the code that have been deemed too buggy. I have just thrown that switch now -- you MUST be running the latest CVS (as of 1130 UTC Sunday) or the other nodes on the network won't let you talk to them (they could still relay requests to you, however). The version check is based on build number, and I have just jacked the build number from 203 where it had been forever, on the experimental branch, to 400 (for 0.4 ;-). There is a jar of build 400 at http://freenetproject.org/fred.jar There are a couple of known problems still: - announcement is a little flaky but will still work; it is recommended that you only have one node ref in your seedNodes file when you do an announcement and things should be fine. - if the node runs a long time it may start refusing all incoming connections saying "connection limit reached" even though there is no activity. this one is a bit of a mystery, so until we figure it out, it is recommended that everyone just restart their node once every day or two. Keep in mind that the message handling code has been redone somewhat so there might be some strange new problems, but overall the new architecture is more robust. IF YOU HAVE A BUG REPORT: - Log on debug, as always. - Turn diagnostics OFF - it generates way too many log messages. - Set doAnnounce=no - it also generates too many log messages (unless of course you have an announcing-related bug). ------------------------------------------------------------------------ I think we have come to a point where we should really get an actual 0.4 release out - pretty soon 0.5 will be ready and it will be too late >;p Where are you Mr. Bad? :-) -- : it's time we took labor saving away from machines : : and gave it back to the people... : _______________________________________________ Chat mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/chat
