Custom Query (1843 matches)

If a request is made for an unresolvable domain, Orchid will repeatedly attempt to connect to the domain until restarted. A request to ​http://foo.org, for example, will fail initially, prompting a site not found error in the browser, and then Orchid will over time repeatedly attempt to connect, indicated in the UI.

If the site is not resolvable, Orchid should register the fact and abort all connection attempts until the site is manually re-requested.

Consider the following code:

         // Peek at head of _outboundQueue and see if we can send it.
            // If so, pull it off, put it in _outbundMessages, test
            // again for bandwidth if necessary, and return it.
            OutboundMessageState state;
            while ((state = _outboundQueue.peek()) != null &&
                   ShouldSend.YES == locked_shouldSend(state)) {
                // we could get a different state, or null, when we poll,
                // due to AQM drops, so we test again if necessary
                OutboundMessageState dequeuedState = _outboundQueue.poll();
                if (dequeuedState != null) {
                    _outboundMessages.add(dequeuedState);
                    if (dequeuedState == state || ShouldSend.YES == locked_shouldSend(state)) {
                        if (_log.shouldLog(Log.DEBUG))
                            _log.debug("Allocate sending (NEW) to " + _remotePeer + ": " + dequeuedState.getMessageId());
                        if (rv == null)
                            rv = new ArrayList<OutboundMessageState>(MAX_ALLOCATE_SEND);
                        rv.add(state);
                        if (rv.size() >= MAX_ALLOCATE_SEND)
                            return rv;
                    }
                }
            }

And note that locked_shouldSend() updates the nextSendTime to be rto from now. So if a higher priority message is inserted in the queue between the peek() and the poll(), the check for equality will fail and the second check will ALWAYS fail (because the first thing locked_shouldSend checks is the nextSendTime).

As a result, the higher priority message will enter the _outboundMessages but the previous head of the queue will not be sent until one rto from now, causing a stall.

The following logs illustrate the effects:

2019/08/01 19:10:49.853 DEBUG [acket pusher] router.transport.udp.PeerState: Nothing to send to [Hash: 2eCs7jCRx6XP7f0tEho-Sv8-sl~6FrmnxkVmA6HpdNs=], with 0 / 5 remaining

create a jdk project for the browser build scripts, which downloads all the jdks from ​https://adoptopenjdk.net/ and prepare them for jlink.