1/20/2016 1:31:05 AM
Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/20/2016 1:29:23 AM
to process all ready notifications before moving the clock.
This should fix https://github.com/killbill/killbill/issues/475.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/19/2016 11:17:30 AM
to https://github.com/killbill/killbill/issues/174.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/19/2016 8:32:42 AM
overdue refresh, InvoicePaymentControlPluginApi callbacks
can take longer now, making tests timeout when Travis is slow.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/18/2016 8:03:46 PM
Meyer <pierre@mouraf.org>
|
1/18/2016 7:27:22 PM
to a race condition, we cannot just rely on payment events to refresh
the overdue state, as the billing state might not be up-to-date yet when
the event is processed.
As a work-around, trigger an overdue refresh after notifying invoice of the payment.
Note that we now take the account lock in overdue, otherwise multiple refreshes would
happen concurrently (the whole overdue process is not atomic, as we modify
entitlements, tags, etc.).
This fixes https://github.com/killbill/killbill/issues/472.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/18/2016 5:10:15 PM
helpful to debug https://github.com/killbill/killbill/issues/472.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/18/2016 12:46:49 PM
Meyer <pierre@mouraf.org>
|
1/18/2016 11:55:22 AM
was sometimes failing due to multiple events being dispatched.
The root cause was the following sequence of events:
* Bus event claimed and dispatched
* Test listener notified
* Other listeners (e.g. invoice) start processing the event
* Clock moved
* Bus re-dispatching the same event (because the claim time has now passed)
The reason the clock is moved too early is because we don't listen to null invoice events in the tests.
We already had a work-around in TestApiListener (we wait until there is no more entries in bus_events), however
one code path was by-passing it. This fixes it -- the main drawback is that we introduce a 500ms delay in the worst
case scenario. We could reduce that arbitrary sleep delay in case it causes a significant slowdown in the tests.
This should fix https://github.com/killbill/killbill/issues/382: I ran the test successfully twice with invocationCount = 100
(while before I could trigger the bug by running the test manually a few times).
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
1/17/2016 8:18:40 PM
fixes https://github.com/killbill/killbill/issues/446.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
|
