On 16.10.14 12:32 , Marcel Reutegger wrote:
Hi,
it doesn't really explain why it happens, but there is an open issue
for the observation tests running on Oak. Some of them fail sporadically:
https://issues.apache.org/jira/browse/JCR-3823
Might be the same root cause...
Might be... However so far I have only seen this on very specific
environments (e.g. Apache buildbot on Windows).
@Bertrand, one subtle change from Jackrabbit 2 is that subsequent calls
to the same EventListener.onEvent call back might happen on different
threads (they won't overlap though). If the EventListener implementation
keeps state across calls, the state from a previous call might not be
visible to a later call if no memory barrier is in place. See
http://docs.oracle.com/javase/7/docs/api/java/util/concurrent/package-summary.html#MemoryVisibility
Michael
Regards
Marcel
On 16/10/14 09:13, "Bertrand Delacretaz" <bdelacre...@apache.org> wrote:
Hi,
I'm sometimes seeing a 10-second delay in delivering observation
events, in Sling integration tests running on Oak 1.07 with TarMK.
Details at SLING-4058.
At this point I'm not sure if Oak or Sling is the culprit as I'm
listening for the OSGi events that Sling generates, so not directly to
Oak events - still investigating.
But if someone has pointers as to what might be wrong, please let me know!
-Bertrand