RFR 8009581: Xpathexception does not honor initcause() (original) (raw)

Aleksej Efimov aleksej.efimov at oracle.com
Fri May 31 13:43:42 UTC 2013

Joe, I'll move the tests to jdk/test/javax/xml/jaxp/XPath/ 8009579 in next version of webrev for JDK8. For JDK7 the tests will be in the same folder after Daniel's change. Correct, I'm using the backport ID only for review in this alias, for commits and test header will use the main bug ID. Thanks for bringing this to my attention.

-Aleksej

On 05/29/2013 03:25 AM, huizhe wang wrote:

Hi Aleksej,

Please note that in jdk8 repo, we've added jdk/test/javax/xml/jaxp to hold all new tests for jaxp. All of the existing tests will be migrated to this place in the future. In your case, the new test can be created under a new folder jdk/test/javax/xml/jaxp/XPath/8009579. Daniel's change request 8008738 will also add that in JDK7 repo., so please coordinate with Daniel when your patch is ready for JDK7. I noticed that you're using the 'backport' bug id for this review. According to the process, all commits shall use the main ID. Thanks, Joe On 5/28/2013 7:30 AM, Aleksej Efimov wrote: Jason, You are absolutely right and my fault that I didn't tested the compatibility of new/old versions. I have tested it with JDK 8(without fix) as a serializer and JDK 8 (with fix) as deserializer. The behavior was as you expected. Thanks for catching it. I'll wait for other comments before generating another webrev.

Aleksej On 05/28/2013 05:41 PM, Jason Mehrens wrote: > Alan, David, > thank you for comments - I also agree with all of them. And as a result v3: > http://cr.openjdk.java.net/~dmeetry/8009581/webrev.3/ > >> I think this looks better. I assume that since the super.getCause() is > >> null that there is no need to handle IllegalStateException now.

You can still generate a ISE inside readObject by performing "new XPathException(new Exception()).initCause(null)" in the old code and then reading that in the new patch. In the new code that line would fail fast. Jason

More information about the core-libs-dev mailing list