[ 
https://issues.apache.org/jira/browse/THRIFT-1805?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15879666#comment-15879666
 ] 

Christopher Tubbs commented on THRIFT-1805:
-------------------------------------------

In Accumulo, I added a test, "ThriftBehaviorIT.java" which uses a 
bi-directional in-memory transport to test exception handling behavior in 
Thrift, using our internal "RpcWrapper" to wrap the service processor and catch 
RuntimeExceptions.
(contained in this commit: 
https://github.com/apache/accumulo/commit/86e6fb44bbf59109a61b159d2bd3bcd0fcfdfa9f)

I'd like the test (or something like it) to be adapted to prevent regressions 
in Thrift directly. I might even be willing to work on it. However, the current 
behavior is undefined when the server encounters a RuntimeException, and 
undefined behavior is untestable. I'm not sure a regression test even makes 
sense unless the most recent regression is also reverted back to a well-defined 
behavior.

> Thrift should not swallow ALL exceptions
> ----------------------------------------
>
>                 Key: THRIFT-1805
>                 URL: https://issues.apache.org/jira/browse/THRIFT-1805
>             Project: Thrift
>          Issue Type: Bug
>          Components: Java - Compiler, Java - Library
>    Affects Versions: 0.9
>            Reporter: Diwaker Gupta
>            Assignee: Diwaker Gupta
>         Attachments: THRIFT-1805.patch
>
>
> In Thrift 0.8.0, Thrift generated Java code did not swallow application 
> exceptions. As a result of THRIFT-1658, this behavior changed in 0.9.0 and 
> now the generated code swallows ALL application exceptions (via 
> ProcessFunction). Apparently this was the behavior in Thrift 0.6.0 and while 
> I see the rationale, it is breaking our applications.
> Our code relies on the fact that exceptions can propagate outside of Thrift 
> for certain things (e.g., to aggressively drop connections for clients that 
> send invalid/malformed requests). ProcessFunction makes it near impossible to 
> do this -- not only does it swallow the exception, it also loses all 
> information about the original exception and just writes out a generic 
> TApplicationException.
> IMO ProcessFunction should only catch TException. If the application code 
> wants to use other exceptions for some reason (in particular, Errors and 
> RuntimeExceptions), Thrift shouldn't prevent that.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Reply via email to