Skip to content
Commits on Jan 18, 2013
  1. 'Release Script: update versions for 5.2.0.CR2'

    Mircea Markus committed Jan 18, 2013
  2. @danberindei

    ISPN-2632 Uneven request balancing after node crash

    Send the cache topology id to the client instead of the JGroups view id.
    After a rebalance, the cache topology id is incremented, so this ensures
    that the client's topology information is updated as well.
    danberindei committed with Mircea Markus Jan 18, 2013
  3. @danberindei

    ISPN-2420 Increment the topology id when a node leaves the cache

    Needed for the ISPN-2632 fix with replicated caches - since they don't
    rebalance after a leave, they will never send a topology update to their
    danberindei committed with Mircea Markus Jan 18, 2013
  4. @anistor

    ISPN-2728 AbstractCacheTransaction.hasLockOrIsLockBackup may cause NPE

    - temporary fix
    Mircea Markus committed with anistor Jan 18, 2013
  5. @anistor

    ISPN-2578 Two PrepareCommands in parallel cause ConcurrentModificatio…

    Mircea Markus committed with anistor Jan 17, 2013
  6. @anistor @danberindei

    ISPN-2366 Rename state transfer config option 'waitForInitialStateTra…

    …nsferToComplete' to 'awaitInitialTransfer'
    * The previous name was not choosen well.
    * Revised javadoc and XSD documentation for this config option to higlight the consequences of turning it off.
    anistor committed with danberindei Jan 18, 2013
  7. @anistor

    ISPN-2405 Pessimistic transactions can leak locks when cluster topolo…

    …gy changes
    * In PessimisticLockingInterceptor.visitLockControlCommand there is an optimization to avoid remote calls if the key is local. Unfortunatelly this had the unintended side effect of not adding the locked key to TX.afectedKeys collection if the key was local which resulted in the key not being unlocked on TX end.
    * BaseRpcInterceptor.shouldInvokeRemoteTxCommand() and AbstractEnlistmentAdapter.mayHaveRemoteLocks will return true if the TX was started in an older topology than current, regardless if modifications or remote locks are present. This ensures the TX command reaches all locked key owners not just the ones registered in TX.getRemoteLocksAcquired() which can become out of date after rehashing.
    * RpcManagerImpl.getTopologyId() should return -1 instead of throwing NPE if the cache has not joined yet (ie. during preloading)
    * Add unit test for the pessimistic locking case.
    * Minor cleanup in the existing UT for optimistic locking.
    anistor committed Oct 15, 2012
  8. @saturnism @galderz
Commits on Jan 17, 2013
  1. @galderz

    ISPN-2632 Added asymmetric cluster tests for Hot Rod server/clients

    * Indirectly fixed the replication queue configuration issue that
    was noticed as part of ISPN-2335.
    galderz committed with Mircea Markus Jan 15, 2013
  2. ISPN-2725 Upgrade to JGroups 3.2.6.Final

    Mircea Markus committed Jan 17, 2013
  3. @galderz

    ISPN-1583 Avoid delegate advanced caches wrapped parameters

    * Such parameters are flags or classloaders.
    * By adding an interface that generates wrap objects, abstract
    delegate implementations can pass on caches with flags or cache
    loaders while keeping these wrapped.
    * This avoid the problem of unwrapping a wrapper while keeping
    hold of the given flag/classloaders.
    galderz committed with Mircea Markus Jan 10, 2013
  4. @galderz

    ISPN-2300 Revert changed flag for entries where conditional ops failed

    * Changed cache entry flagged was overloaded to mean that both that a
    new entry had to be stored in the cache, and that a entry loaded from
    the cache store had to be stored in the cache.
    * Changed now represents a change, so a cache entry whose value in
    memory is going to change as a direct result of the cache operation,
    i.e. when the conditional operation succeeds.
    * Loaded means that a entry was loaded from the cache store and needs
    to be stored in the cache, i.e. when a conditional operation requires
    comparing the old value which is present in the cache store, and if
    the condition fails, the value in the cache store still needs updating
    in the cache.
    galderz committed with Mircea Markus Jan 16, 2013
  5. @vblagoje @galderz

    ISPN-2439 Deadlock in Map/Reduce tasks

    * Make sure the cluster is started before using the intermediate cache
    vblagoje committed with galderz Oct 30, 2012
  6. @tristantarrant @galderz
  7. @tristantarrant @galderz
  8. @tristantarrant @galderz

    ISPN-2545 Explicitly add the slf4j-log4j12 implementation to the test…

    … dependencies to avoid issues with IBM JDK
    tristantarrant committed with galderz Jan 15, 2013
  9. @tristantarrant @galderz

    ISPN-2706 Minor help fixes

    tristantarrant committed with galderz Jan 11, 2013
  10. @tristantarrant @galderz
Commits on Jan 16, 2013
  1. @anistor @galderz

    ISPN-2711 NPE in StateConsumerImpl.addUpdatedKey and isKeyUpdated

    updatedKeys can be set to null by another thread that calls stopApplyingState() between the initial null check and the actual access leading to NPE. To guard against this it is enough to grab a local copy of the reference.
    anistor committed with galderz Jan 15, 2013
  2. @anistor @galderz

    ISPN-2326 c3p0 not part of the release

    * c3p0 dependency was ignored by maven assembly plugin because this dependency is referenced by multiple modules, in multiple scopes (including test scope). The processing order of modules (and their dependencies) becomes important because only the first one is considered, so if that happens to have 'test' scope it won't make it into the distro.  This is a consequence of a broader issue/limitation that is already documented in the plugin user guide and to overcome it we should avoid multiple moduleSets (not doable). The current solution consists of adding a FIRST moduleSet that aggregates all jar deps in 'proper' order and removing dependency processing from existing moduleSets.
    * Made some trivial changes in bin.xml assembly to reduce the amount of text differences to all.xml (bin.xml is a subset of all.xml and any irrelevant differences should go away to simplify maintaining them in parallel)
    * Force dependency resolution for commons-codec and commons-collections to use a global common version (otherwise they would diverge)
    * Remove some duplicated dependencies in spring/pom.xml and  cli-server/pom.xml.
    anistor committed with galderz Jan 14, 2013
Commits on Jan 11, 2013
  1. @galderz
  2. @galderz

    ISPN-2700 Fix Scala compiler warnings

    galderz committed with Mircea Markus Jan 9, 2013
Commits on Jan 10, 2013
  1. ISPN-2690: Removed Serializable and updatd internal externalizer tabl…

    …e to reference CHF Externalizer implementations.
    Erik Salter committed Jan 10, 2013
  2. @galderz @tristantarrant

    ISPN-2054 isPre=false events now happen *after* cache has been updated

    * Before this fix, isPre=false event notifications had no guarantees
    that the events were received after cache had been updated, leading
    to potential issues with cache state assumptions.
    * This now has been fixed for cache entry created, modified, evicted
    and removed events.
    galderz committed with tristantarrant Dec 20, 2012
  3. @danberindei @galderz

    ISPN-2701 Disable ConfigurationOverridesTest, InfinispanContainerCach…

    …eManagerDefinitionTest and InfinispanEmbeddedCacheManagerDefinitionTest temporarily.
    danberindei committed with galderz Jan 8, 2013
  4. @tristantarrant

    ISPN-2660 Replace non-alphanumeric characters in cache names with an …

    …underscore to prevent database naming errors
    tristantarrant committed Jan 9, 2013
Commits on Jan 8, 2013
  1. fixed RemoteCacheStoreTest

    Mircea Markus committed Jan 8, 2013
  2. @galderz

    ISPN-2696 Upgrade to Scala 2.10.0

    galderz committed with Mircea Markus Jan 8, 2013
  3. @anistor

    ISPN-2698 Some in-progress segment transfers are not cancelled proper…

    …ly during rehashing
    StateConsumerImpl.cancelTransfers(..) was ignoring the currently iterated segment.
    anistor committed with Mircea Markus Jan 8, 2013
  4. @anistor

    ISPN-2580 Do not request segments from all nodes at once

    * Add InboundTransferTask.awaitCompletion() to be able to detect when all segments from a source were completely received
    * StateConsumerImpl.applyState processes all the received chunks at once
    * StateConsumerImpl.onTopologyUpdate() no longer creates InboundTransferTasks if cache is not transactional and fetching of in-memory state is not enabled. These were previosly created only to be removed.
    * StateConsumerImpl.addTransfers() is split into separate methods for fetching transactions and segments
    * StateConsumerImpl requests transactions serially from neighbours and then notifies that transaction data was received without waiting for the segments to be requested. A separate task is spawned to request transactions serially from neighbours.
    anistor committed with Mircea Markus Jan 3, 2013
  5. @anistor

    Improve asserts.

    anistor committed with Mircea Markus Jan 3, 2013
  6. @anistor

    Fix generated node name for indices higher than 'Z'.

    Before the fix is was possible that the generated name contained non-printable characters. The current implementation uses multiple letters if the index is higher than 'Z'.
    anistor committed with Mircea Markus Jan 3, 2013
  7. @anistor

    Fix exception message.

    anistor committed with Mircea Markus Jan 3, 2013
  8. @anistor

    Fix TestNG test name to match class name.

    anistor committed with Mircea Markus Jan 3, 2013
Something went wrong with that request. Please try again.