Changes
1.0.5
- fixed a race condition that would occur if a user (or the failover extension)
tried to re-connect an OnStomp::Client instance within an
on_connection_closed
(or similar) event handler. The fix is more of a band-aid at the moment,
and a proper fix will follow. Thanks to alcy for
helping me track this down.
1.0.4
- fixed major oversight when using ruby 1.8.7 / jruby with an SSL connection.
it will use blocking read/write methods if nonblocking methods are unavailable
- added write and read block timeouts. if there is no write activity for a period
of time (default of 120 seconds), the connection is assumed to be dead. read
activity is handled similarly, but only when performing the CONNECT/CONNECTED
exchange, after that we have no way of knowing if the broker just doesn't
have any data for us (unless using a STOMP 1.1 connection with heartbeating)
- refactored common failover buffer code into lib/onstomp/failover/buffers/base
1.0.3
- change how failover spawns new connections when failing over.
1.0.2
- allow failover clients to be constructed from regular OnStomp::Client
instances, allowing fully configured SSL connections.
1.0.1
- improved failover buffer handling
- added RECEIPT driven failover buffer handling - more reliable, but slower
as it requests receipts for nearly all frames.
1.0.0
- initial release
- support for STOMP 1.0 and STOMP 1.1 protocols
- fully featured client with event bindings and non-blocking IO
- experimental support for open-uri STOMP interface
- experimental support for failover/reliable connection handling
1.0pre - 2011-03-30