Joe McDonnell created IMPALA-7461:
-------------------------------------
Summary: TestDecimalFuzz hits DCHECK math-functions-ir.cc "Check
failed lhs > 0 && rhs > 0"
Key: IMPALA-7461
URL: https://issues.apache.org/jira/browse/IMPALA-7461
Project: IMPALA
Issue Type: Bug
Components: Backend
Affects Versions: Impala 3.1.0
Reporter: Joe McDonnell
Several builds have been failing with an impalad hitting this DCHECK:
{code:java}
auto MultiplicationOverflows = [](ActualType lhs, ActualType rhs) {
DCHECK(lhs > 0 && rhs > 0); <----- HERE
using ActualType = decltype(lhs);
return BitUtil::CountLeadingZeros(lhs) + BitUtil::CountLeadingZeros(rhs) <=
BitUtil::UnsignedWidth<ActualType>() + 1;
};
{code}
This happens during the TestDecimalFuzz test:
{noformat}
I0817 10:50:44.950784 10364 Frontend.java:1024] Analyzing query: select
width_bucket(cast(-10.0 as decimal(8,6)), cast(-10.0 as decimal(3,1)),
cast(6490371073.16853453566312041152511 as decimal(36,25)), 26386065)
F0817 10:50:44.952258 10364 math-functions-ir.cc:516] Check failed: lhs > 0 &&
rhs > 0{noformat}
It looks like a regression due to the fix for IMPALA-7412.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
