[infinispan-dev] 5.3 tests to confirm users' API view of tx isolation choice?

[cotton-ben]

Hi,

FYI.  I am trying to get a test to run on 5.3.0.a1 to confirm that
DIRTY_READ intolerant transactional use cases are accommodated by the
Infinispan JSR-107 API 0.4 implementation.

There are 2 programs attached.  

The first one does not use JSR-107 at all, but rather mocks an example of a
program that  exhibits the trauma (inconsistent data!)  of enduring a
DIRTY_READ event (i.e. TX_THREAD_2 does not block @t=2).  The DIRTY_READ
endured in this program is cruelly coerced ... exclusively for the purpose
of demonstrating the @t=X events.

The second one borrows from the first program and explicitly uses the
JSR-107 API 0.4 (JCACHE provider=Infinispan 5.3.0.a1) to specify its
DIRTY_READ intolerance.  This program has the expectation that Infinispan
5.3.0.a1 can be configured with a lockPolicy=PESSIMISTIC so that TX_THREAD_2
*blocks* @t=2).

*I have not been able to get the second program to work w/ Infinispan
5.3.0.a1 as a JCACHE 0.4 provider.*  

Please consider helping to get the second program running, as I believe it
exercise what should be a minimum expectation of what a JCACHE end-user will
have for an Infinispan implementation that delivers the JSR-107 transactions
option.

program 1 (show what the test wants to exercise) ==> 
_DirtyReadIntolerantDriver.java
<http://infinispan-developer-list.980875.n3.nabble.com/file/n4026949/_DirtyReadIntolerantDriver.java>  

program 2 (exercise the test on Infinispan 5.3.0a1's implementation of
JCACHE 0.4 API) ==>  DirtyReadIntolerantDriver.java
<http://infinispan-developer-list.980875.n3.nabble.com/file/n4026949/DirtyReadIntolerantDriver.java>  

I know all of you are busy with priority devl matters, thanks for any
consideration to help with this request.

Ben








--
View this message in context: http://infinispan-developer-list.980875.n3.nabble.com/5-3-tests-to-confirm-users-API-view-of-tx-isolation-choice-tp4026942p4026949.html
Sent from the Infinispan Developer List mailing list archive at Nabble.com.