I've made those changes and they seem to resolve the issue for me both on
Java 6 and Java 7.

Will commit the change shortly

Rob

On 26/12/2013 11:56, "Andy Seaborne" <[email protected]> wrote:

>Rob,
>
>I can't reliably run the JDBC tests reliably on my local machine but I
>think I know what's going on.
>
>The tests freeze up and the suite never finishes at
>oaj.jdbc.mem.connections.TestDatasetConnection.
>
>When I get it it fail, it is always in this test:
>AbstractJenaConnectionTests.connection_statement_query_construct_timeout_0
>2()
>
>
>It happens with java7 with maven from the command line but not java6
>(OpenJDK or sun). In Eclipse, with Java6, it can lock up as well.  The
>version of maven does not affect it.  I am not seeing the problem we had
>before where there were huge numbers of TCP connections waiting to get
>cleaned up by the kernel.  There is no significant CPU being used.
>
>Reducing the test data size and it terminates with "no exception" as
>expected.  As set up, I think it will fail after 30 hours but I haven't
>waited that long.
>
>Poking at the test, adding debug statements, I have seen the
>SQLException only occur at point of stmt.executeQuery, and the test
>works OK.  I have not see an SQLException from the rset.next() loop.
>
>What seems to be happening is that the query executes quite quickly
>sometimes, no timeout, and then the results materialized by
>JenaStatement (it's of type ResultSet.TYPE_SCROLL_INSENSITIVE).
>
>After this, there is no opportunity for the query to timeout - it's
>finished.  Looping on rset.next is not pulling results because in this
>test, the result are materialized at the start.
>
>Possible fix:
>1/ Add a delay FILTER(<http://jena.hpl.hp.com/ARQ/function#wait>(1000))
>to make sure the query is slow enough.
>
>2/ Consider what this test is actually testing because the results are
>materialised by stmt.executeQuery when ResultSet.TYPE_SCROLL_INSENSITIVE.
>
>See also connection_statement_query_select_timeout_02
>which has a 3-way cross product, not a 2-way one.
>
>
>       Andy




Reply via email to