On 26/06/13 11:42, Andy Seaborne wrote:
On 26/06/13 00:07, Rob Vesse wrote:
Can this commit be rolled back in some way?
Sure - seems reasonable.
Might be a good it as graceful server shutdown, which includes
deregistration of datasets.
I'll take a look ASAP.
JENA-478
It hampers test isolation because the server isn't a unique instance each
time. Would a better fix not be to deregister JMX in the stopServer()
method?
Rob -
Looking at ARQMgt, JMX beans get deregistered if they are already
registered so (1) this is broken or (2) some thing else is going on.
The only issue I can see is that ServerTest has a class static
setupServer() ; which creates an initial server that hangs around. If I
remove that, servers start and stop according to the test class.
Usually that's
@BeforeClass public static void beforeClass()
{ ServerTest.allocServer() ; }
@AfterClass public static void afterClass()
{ ServerTest.freeServer() ; }
@Before public void before()
{ ServerTest.resetServer() ; }
but you could @before/@after for each test. Server setup is not fast.
Andy
Something somewhere is printing to stderr:
LoginService=HashLoginService[Fuseki Authentication]
identityService=org.eclipse.jetty.security.DefaultIdentityService@33c4bcda
This may be jetty - an upgrade of versions may fix there. There were a
few of these in a recent version.
Andy
