Permalink
Commits on May 30, 2017
Commits on May 26, 2017
Commits on May 18, 2017
Commits on May 4, 2017
  1. Auto-close async HEAD requests after the first send [IMMUTANT-630]

    tobias committed May 4, 2017
    If a HEAD request is handled by something that creates an async channel,
    the response should use the same headers and status as a GET
    would. However, if the user detects the HEAD and returns an empty body,
    the response will have a 'Content-Length: 0' header instead of the
    'Transfer-Encoding: Chunked' header that a GET would get, which is a
    violation of the HTTP
    spec (https://tools.ietf.org/html/rfc7230#section-3.3.2). And if we let
    the regular channel semantics occur, we may leak connections since they
    may not get closed (and there's no value in keeping them open beyond the
    first send, since the body of HEAD responses are ignored).
Commits on Jun 27, 2016
Commits on Apr 20, 2016
Commits on Apr 15, 2016
  1. Attach the websocket handshake handler to the outer handler [IMMUTANT…

    tobias committed Apr 15, 2016
    …-616]
    
    This allows users to access the handler in order to manipulate it before
    it gets used (to add extensions, for example).
Commits on Mar 8, 2016
  1. Revert back to Undertow 1.3.0.Beta9

    tobias committed Mar 8, 2016
    1.3.18.Final has some performance regressions that the Undertow team is
    investigating, so we'll stick with the version we used in the last
    release until they figure out a fix.
Commits on Jan 11, 2016
  1. Update copyrights.

    tobias committed Jan 11, 2016
Commits on Jan 8, 2016
Commits on Nov 19, 2015
  1. Don't allow /-segmented dest names to shadow [IMMUTANT-597]

    tobias committed Nov 19, 2015
    If we get back a NamingContext, we pretend we got nothing.
Commits on Nov 11, 2015
Commits on Nov 10, 2015
  1. Use a concurrency of 1 for response routers [IMMUTANT-593]

    tobias committed Nov 10, 2015
    Using numProcs is wasteful, since the listener just satisfies a future.
Commits on Nov 6, 2015
  1. Use the simple election policy for singletons when in WF 8/9

    tobias committed Nov 6, 2015
    There is a bug (https://issues.jboss.org/browse/WFLY-5108) that can
    cause non-deterministic election policies to result in an inconsistent
    singleton state (multiple nodes running the singleton, or no nodes
    running the singleton are both possible). This caused an intermittent
    integ failure in Immutant's cluster tests. We're still using the random
    election policy with WF10 - the Immutant tests should catch it if it
    also has the same behavior.
Commits on Nov 5, 2015
  1. Invoke the WS init handler before the socket is connected

    tobias committed Nov 5, 2015
    Otherwise, the init handler can't set headers, since the headers have
    been written by the time the socket is connected.
    
    This change also lets us get rid of that silly threadLocal hack.
Commits on Nov 3, 2015
Commits on Oct 29, 2015
  1. Explicitly depend on WF10's messaging-activemq extension

    tobias committed Oct 29, 2015
    We need this for messaging to work in wildfly-swarm.
    
    This also reorgs the file a bit to group related modules together.
  2. Detect if we are inside wildfly-swarm

    tobias committed Oct 29, 2015
    And treat it as WildFly 10.
Commits on Oct 22, 2015
  1. Adjust CI script to new path

    tobias committed Oct 22, 2015
  2. Remove unneeded module dir

    tobias committed Oct 22, 2015
    This moves all of the modules to the top-level, removing the extra dir &
    pom.
  3. Break wildfly up in to multiple modules

    tobias committed Oct 22, 2015
    This is the first step in allowing wunderboss-based apps to properly
    take advantage of swarm, since wildfly functionality can be brought in
    piecemeal, instead of as one dep. This also means that we don't assume a
    full WF is available, and each piece only assumes what it needs is
    there.
    
    There is still an org.projectodd.wunderboss/wunderboss-wildfly, but it
    is an aggregator for all of the other WF modules.
Commits on Oct 9, 2015
  1. Use WildFly's built-in singleton support.

    tobias committed Oct 7, 2015
    This greatly simplifies our own singleton implementation, and makes it
    more reliable, in theory.
Commits on Oct 7, 2015
Commits on Oct 6, 2015
  1. Notify stop callback whenever stop is called

    tobias committed Oct 6, 2015
    even if the daemon is no longer running.