[jira] [Assigned] (CAMEL-4022) Issue using errorBuilderRef with the xml dsl

["Claus Ibsen (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CAMEL-4022?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Claus Ibsen reassigned CAMEL-4022:
----------------------------------

    Assignee: Claus Ibsen  (was: Hadrian Zbarcea)

> Issue using errorBuilderRef with the xml dsl
> --------------------------------------------
>
>                 Key: CAMEL-4022
>                 URL: https://issues.apache.org/jira/browse/CAMEL-4022
>             Project: Camel
>          Issue Type: Bug
>    Affects Versions: 2.7.1
>            Reporter: Hadrian Zbarcea
>            Assignee: Claus Ibsen
>            Priority: Critical
>         Attachments: onexceptionerrorhandler.patch
>
>
> While fixing issues around the errorHandler I noticed that <onException> definitions defined in the camel context are ignored if a route specifies its own errorHandlerRef. The reason is that we set the onException definition on the default error handler. I have a fix for that, but I discovered a different issue (I think) for which I would like to discuss the solution.
> When we have an onException definition that looks kinda like this:
> {code}
> <onException>
>   <exception> java.lang.IllegalArgumentException</exception>
>   <to uri="mock:illegalArgumentException"/>
> </onException>
> {code}
> ... something happens, the IAE exception is caught, we do something, but in that process another exception is thrown. Currently, that would be caught by the default error handler, which may not be what we want.
> What error handler (if any) should handle exceptions thrown while in onException?
> The onException mechanism is somewhat similar to a try/catch. I don't think the exceptions thrown while handling onException should be handled by the same error handler configured for the route, or even the context scoped one. The processing should be very simple, predictable and immutable. Since the default "CamelDefaultErrorHandlerBuilder" can be replaced, it is not imho a solution and we need one global one that does as little as possible (the problem would be agreeing what that is: no redeliveries, logging or not, etc).
> Thoughts?

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira