[jboss-jira] [JBoss JIRA] (DROOLS-5284) MVELTest.testTypeCoercionLongDivByInt fails with IBM JDK 8

[Toshiya Kobayashi (Jira)]

    [ https://issues.redhat.com/browse/DROOLS-5284?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14067542#comment-14067542 ] 

Toshiya Kobayashi edited comment on DROOLS-5284 at 4/30/20 3:45 AM:
--------------------------------------------------------------------

First of all, the expression in the unit test is not good. Java cannot compile this because Math.round() can accept only double or float.

{code:java}
15 * Math.round( new BigDecimal("49.4") ) / 100
{code}

Then, MVEL kindly tries to find a "BestCandidate" method for this.

https://github.com/mvel/mvel/blob/mvel2-2.4.7.Final/src/main/java/org/mvel2/util/ParseTools.java#L232

So MVEL picks up

{code:java}
public static int java.lang.Math.round(float)
{code}

or

{code:java}
public static long java.lang.Math.round(double)
{code}

(Assuming MVEL can coerce BigDecimal to float or double).

However, these 2 methods have no superiority in terms of scoring for "BestCandidate". Both get score "4".

https://github.com/mvel/mvel/blob/mvel2-2.4.7.Final/src/main/java/org/mvel2/util/ParseTools.java#L312

So the result depends on which method was picked first from a method list (The method list is given by Class.getMethods()).

In case of OpenJDK, Math.round(double) is picked. -> Finally, the expression results in (Long / Long) which returns Double by MathProcessor.

In case of IBM JDK Math.round(float) is picked. -> Finally, the expression results in (Integer / Integer) which returns Integer by IntDiv.

I think I can tweak the selection logic as Double is preferred coercion for BigDecimal.


was (Author: tkobayashi):
First of all, the expression in the unit test is not good. Java cannot compile this because Math.round() can accept only double or float.

{code:java}
15 * Math.round( new BigDecimal("49.4") ) / 100
{code}

Then, MVEL kindly tries to find a "BestCandidate" method for this.

https://github.com/mvel/mvel/blob/mvel2-2.4.7.Final/src/main/java/org/mvel2/util/ParseTools.java#L232

So MVEL picks up

{code:java}
public static int java.lang.Math.round(float)
{code}

or

{code:java}
public static long java.lang.Math.round(double)
{code}

(Assuming MVEL can coerce BigDecimal to float or double).

However, these 2 methods has no superiority in terms of scoring for "BestCandidate". Both get score "4".

https://github.com/mvel/mvel/blob/mvel2-2.4.7.Final/src/main/java/org/mvel2/util/ParseTools.java#L312

So the result depends on which method was picked first (The method lists are given by Class.getMethods()).

In case of OpenJDK, Math.round(double) is picked. -> Finally, the expression results in (Long / Long) which returns Double by MathProcessor.

In case of IBM JDK Math.round(float) is picked. -> Finally, the expression results in (Integer / Integer) which returns Integer by IntDiv.

I think I can tweak the selection logic as Double is preferred coercion for BigDecimal.

> MVELTest.testTypeCoercionLongDivByInt fails with IBM JDK 8
> ----------------------------------------------------------
>
>                 Key: DROOLS-5284
>                 URL: https://issues.redhat.com/browse/DROOLS-5284
>             Project: Drools
>          Issue Type: Bug
>          Components: core engine
>    Affects Versions: 7.36.0.Final
>            Reporter: Toshiya Kobayashi
>            Assignee: Toshiya Kobayashi
>            Priority: Major
>
> MVELTest.testTypeCoercionLongDivByInt fails with IBM JDK 8
> {noformat}
> [ERROR] Tests run: 1, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.97 s <<< FAILURE! - in org.drools.compiler.integrationtests.MVELTest
> [ERROR] testTypeCoercionLongDivByInt(org.drools.compiler.integrationtests.MVELTest)  Time elapsed: 0.962 s  <<< FAILURE!
> java.lang.AssertionError: expected:<7.350000> but was:<7>
> 	at org.junit.Assert.fail(Assert.java:88)
> 	at org.junit.Assert.failNotEquals(Assert.java:834)
> 	at org.junit.Assert.assertEquals(Assert.java:118)
> 	at org.junit.Assert.assertEquals(Assert.java:144)
> 	at org.drools.compiler.integrationtests.MVELTest.testTypeCoercionLongDivByInt(MVELTest.java:1150)
> 	...
> [INFO] 
> [INFO] Results:
> [INFO] 
> [ERROR] Failures: 
> [ERROR]   MVELTest.testTypeCoercionLongDivByInt:1150 expected:<7.350000> but was:<7>
> {noformat}



--
This message was sent by Atlassian Jira
(v7.13.8#713008)