Mailing-List: contact dev-help@commons.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Commons Developers List" <dev@commons.apache.org>
Received-SPF: neutral (athena.apache.org: local policy)
Message-ID: <4D768EE6.6050707@free.fr>
Date: Tue, 08 Mar 2011 21:17:42 +0100
From: Luc Maisonobe <Luc.Maisonobe@free.fr>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
 rv:1.9.2.14) Gecko/20110223 Thunderbird/3.1.8
MIME-Version: 1.0
To: dev@commons.apache.org
Subject: [math] Re: svn commit: r1078734
References: <20110307103754.1331923889EB@eris.apache.org>
In-Reply-To: <20110307103754.1331923889EB@eris.apache.org>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Le 07/03/2011 11:37, erans@apache.org a écrit :
> Author: erans
> Date: Mon Mar  7 10:37:52 2011
> New Revision: 1078734
> 
> URL: http://svn.apache.org/viewvc?rev=1078734&view=rev
> Log:
> MATH-542
> "MathRuntimeException" provides two ways for enhancing the information content:
> one for localized messages and one for storing "context" objects.
> The additional methods have been added to "MathThrowable". Consequently, dummy
> implementations were needed in the old "MathRuntimeException" and
> "MathException".
> Some parts of the tests for old exceptions were removed as they used methods
> that were deleted. A call to "MathUserException" in "AbstractContinuousDistribution"
> was simplified for the same reason.
> "MathUtils" still contained "String" (instead of "Localizable") as argument to
> the constructor of a "MathArithmeticException".
> I don't know why a test in "DummyStepInterpolatorTest" stopped working. It is now
> commented out; please have a look.

[snip]


> Modified: commons/proper/math/trunk/src/test/java/org/apache/commons/math/ode/sampling/DummyStepInterpolatorTest.java
> URL: http://svn.apache.org/viewvc/commons/proper/math/trunk/src/test/java/org/apache/commons/math/ode/sampling/DummyStepInterpolatorTest.java?rev=1078734&r1=1078733&r2=1078734&view=diff
> ==============================================================================
> --- commons/proper/math/trunk/src/test/java/org/apache/commons/math/ode/sampling/DummyStepInterpolatorTest.java (original)
> +++ commons/proper/math/trunk/src/test/java/org/apache/commons/math/ode/sampling/DummyStepInterpolatorTest.java Mon Mar  7 10:37:52 2011
> @@ -123,7 +123,9 @@ public class DummyStepInterpolatorTest {
>          fail("an exception should have been thrown");
>      } catch (IOException ioe) {
>          // expected behavior
> -        assertEquals(0, ioe.getMessage().length());
> +        // XXX Why was the message supposed to be empty?
> +        // With the current code it is "org.apache.commons.math.util.Pair".
> +        // assertEquals(0, ioe.getMessage().length());

The problem here is that the exception thrown which is an IOException
built from the empty string in the BadStepInterpolator below is not the
IOException that is caught here. In fact the IOException caught is
really a java.io.NotSerializableException which is built with the
message "org.apache.commons.math.util.Pair" automatically by the
serialization process in the JVM (at least for Oracle implementation).

All exceptions must be serializable (this comes from java.lang.Throwable
implementing the Serializable interface). This was also one reason why
our Localizable interface had to extends Serializable, so it could be
used as a field in our exceptions.

This change added a List<Pair<Localizable, Object[]>> field in
MathRuntimeException. Pair is not serializable. Adding Serializable to
the implemented interfaces in Pair does solve the problem (if also the
change below is reverted, of course) but I don't think it would be wise
unless we also enforce the two generic parameters of Pair to be
serializable too. Perhaps we should use a Map<Localizable, Object[]>
instead ?

The problem also makes me think that we already had a similar bug in
another part of the exceptions for a long time : the Object or Object[]
parameters of the exceptions are stored and may not be Serializable too.
This was never a problem either because the parameters are often simple
primitive ones (int, double ...) or arrays so they were serializable.
Perhaps we should change the signature from Object and Object[] to
Serializable and Serializable[] ?


>      }
>  
>    }
> @@ -137,7 +139,7 @@ public class DummyStepInterpolatorTest {
>        }
>        @Override
>        protected void doFinalize() throws MathUserException {
> -          throw new MathUserException((Localizable) null, LocalizedFormats.SIMPLE_MESSAGE, "");
> +          throw new MathUserException(LocalizedFormats.SIMPLE_MESSAGE, null);

This should not have been replaced. Null is not the smae thing as the
empty String. Also this introduces a warning due to ambiguous signatures.

Luc

>        }
>    }


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org