[
https://issues.apache.org/jira/browse/IMPALA-4233?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16840843#comment-16840843
]
Tim Armstrong commented on IMPALA-4233:
---------------------------------------
I think at the moment there's no "expected" reason why codegen might fail,
except for a bug or an unexpected runtime error, so I'm not sure that this
would really achieve that much. I'll close for now.
> Revisit handling of expr codegen failures
> -----------------------------------------
>
> Key: IMPALA-4233
> URL: https://issues.apache.org/jira/browse/IMPALA-4233
> Project: IMPALA
> Issue Type: Improvement
> Components: Backend
> Affects Versions: Impala 2.8.0
> Reporter: bharath v
> Assignee: Michael Ho
> Priority: Minor
> Labels: codegen
>
> FragmentInstanceState::Open() can fail if codegen of expressions fails:
> {code}
> // It shouldn't be fatal to fail codegen. However, until IMPALA-4233 is
> fixed,
> // ScalarFnCall has no fall back to interpretation when codegen fails
> so propagates
> // the error status for now.
> RETURN_IF_ERROR(runtime_state_->CodegenScalarFns());
> {code}
> In principle many functions can be executed without codegen - after
> IMPALA-4356 was merged, we always initialise the interpreted path, unless the
> function *requires* codegen.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
