Permalink
Commits on Mar 8, 2016
  1. Version 2.1.4-SNAPSHOT

    tobias committed Mar 8, 2016
  2. Version 2.1.3

    tobias committed Mar 8, 2016
  3. Use wunderboss release

    tobias committed Mar 8, 2016
Commits on Mar 4, 2016
  1. Dispatch if given a non-String ring body [IMMUTANT-611]

    Instead of disallowing users to use InputStreams or Files from a
    non-dispatched handler, we instead force dispatch just before writing
    the response for those two types, as well as for seqs.
    tobias committed Mar 4, 2016
Commits on Mar 3, 2016
Commits on Mar 2, 2016
  1. Deprecate app-root and app-relative [IMMUTANT-606]

    app-root and app-relative are broken for non-dev wars in-container, and
    we don't have a good way to fix them. Users should be using io/resource
    anyway.
    tobias committed Mar 2, 2016
  2. Merge pull request #23 from ddellacosta/master

    provide a non-decoded :path-info to ring
    tobias committed Mar 2, 2016
  3. Adds path-info' in immutant.web.internal.undertow/.servlet to provide…

    … a non-URL-decoded :path-info value in the Ring request map. [IMMUTANT-610]
    ddellacosta committed Mar 1, 2016
  4. Use an undertow Sender when not dispatching [IMMUTANT-611]

    Undertow now checks that you aren't doing blocking IO when you don't
    dispatch to the worker pool, so without this change any non-dispatched
    handler will fail. Using the non-blocking Sender should also improve the
    performance a bit when not dispatching.
    tobias committed Mar 2, 2016
  5. Return :port and :host from undertow/options [IMMUTANT-612]

    Because it's difficult to tell exactly which listeners are registered
    with the Undertow/Builder associated with the :configuration key, not to
    mention that there may be multiple ports configured, the best we can do
    I think is to return the most recently passed port. It's really either
    that or return no port at all, which for the common case -- one port --
    isn't really helpful -- especially when the port is 0.
    
    Regardless, returning the default port, 8080, after Undertow-specific
    options are passed is just dead wrong, so we fixed that, at least.
    jcrossley3 committed Mar 2, 2016
Commits on Feb 22, 2016
  1. Minor doc cleanup

    jcrossley3 committed Feb 22, 2016
  2. Use at-exit's new shutdown hook [IMMUTANT-605]

    Confirmed that the method calls on the GracefulShutdownHandler are
    idempotent, so no problem if the server is stopped prior to shutdown. If
    a user wants to drain requests after stopping but prior to shutdown,
    they must invoke the methods manually (on the object returned from
    graceful-shutdown, which they ostensibly passed to web/run)
    
    I considered adding an option to web/run for a "drain timeout" and
    always wrapping the handler in a GracefulShutdownHandler. But since this
    is only relevant for Undertow itself, i.e. out-of-container, I thought
    it best to keep the timeout as a parameter to a separate function.
    jcrossley3 committed Feb 22, 2016
Commits on Feb 19, 2016
  1. 1st pass at a shutdown hook idea [IMMUTANT-605]

    Subject to change. Not sure addShutdownHook is valid in all
    environments. Also would like to consider wrapping all HttpHandlers in a
    GracefulShutdownHandler, possibly in wboss/mount.
    jcrossley3 committed Feb 19, 2016
  2. Refactor HttpHandler creation, expose via public ns [IMMUTANT-605]

    New method is now immutant.web.undertow/http-handler which, for now,
    delegates to its internal cousin, which is subject to change.
    jcrossley3 committed Feb 19, 2016
  3. Test against WildFly 10.0.0.Final

    tobias committed Feb 19, 2016
  4. Don't always assume websockets are available [IMMUTANT-600]

    We blindly assumed the javax.websocket.server.ServerContainer was always
    available, but it will only be available in EAP6 if websockets have been
    configured in standalone.xml *and* in a jboss-web.xml in the WAR.
    
    We can't know if a user's code will use websockets, so we have to try to
    set up the endpoint for every servlet deployed. That means that users
    that had no interest in websockets (and hadn't configured them) would
    still be unable to run any ring handlers.
    
    This changes the servlet init code to confirm the ServerContainer is
    available, logging if it isn't found instead of throwing an NPE.
    tobias committed Feb 19, 2016
Commits on Feb 2, 2016
  1. Restore broken-but-possibly-relied-upon behavior [IMMUTANT-604]

    One may once again pass messaging properties as keywords and we'll
    automatically convert them to strings for you.
    jcrossley3 committed Feb 2, 2016
  2. No longer recursively munging option map keys [IMMUTANT-604]

    I'm not sure why we ever thought this was a good idea, probably just
    never thought any option values would be maps. But now they are, and we
    really shouldn't be messing with them.
    
    One bit of fallout from this is message metadata, where keyword keys
    used to be turned into Strings (as they must be valid Java identifiers
    to be used in selectors) automatically. Now they're not. An error will
    occur if you attempt to pass keyword keys as metadata. I think this
    probably better satisfies the "principle of least surprise"
    jcrossley3 committed Feb 2, 2016
Commits on Jan 11, 2016
  1. update copyrights.

    tobias committed Jan 11, 2016
  2. Test against clojure 1.8.0-RC5

    tobias committed Jan 11, 2016
Commits on Jan 9, 2016
  1. Merge pull request #21 from matthiasn/patch-2

    Function name formatting
    tobias committed Jan 9, 2016
  2. Update web.md

    matthiasn committed Jan 9, 2016
  3. Merge pull request #20 from matthiasn/patch-1

    fix typo
    jcrossley3 committed Jan 9, 2016
  4. fix typo

    matthiasn committed Jan 9, 2016
Commits on Jan 8, 2016
  1. Version 2.1.3-SNAPSHOT

    tobias committed Jan 8, 2016
  2. Version 2.1.2

    tobias committed Jan 8, 2016
  3. Use wunderboss release.

    tobias committed Jan 8, 2016
Commits on Dec 16, 2015
  1. Test with clojure 1.8.0-RC4

    tobias committed Dec 16, 2015