RFR (jaxp): 8016133 : Regression: diff. behavior with user-defined SAXParser (original) (raw)

huizhe wang huizhe.wang at oracle.com
Mon Jun 17 16:10:18 UTC 2013

Thanks Chris, Daniel.

Daniel, I'm not a reviewer either. I saw some changesets had my user id on the reviewer list and surprised that the gate keeper didn't complain. Maybe it would if there hasn't been a real reviewer like Chris who's also on the list :-) So in this case, it won't complain a review list like "chegar, dfuchs". It's good for the record, since I know you've reviewed many of my change requests and given great comments.

-Joe

On 6/17/2013 3:42 AM, Chris Hegarty wrote:

Looks ok to me too Joe.

-Chris. On 17/06/2013 09:17, Daniel Fuchs wrote: Hi Joe,

Looks good to me. Not a reviewer - as you know ;-) -- daniel On 6/11/13 5:36 AM, huizhe wang wrote: Hi,

This is a quick fix on a regression caused by a previous patch. The XMLReaderFactory uses a class variable (jarread) to indicate if service file has already been read. Along with this variable, there was another (clsFromJar ) that caches the classname if found in a service file. The 2nd variable and its use were accidentally removed. As a result, the following code would return the 3rd party impl on first call but then fall back to the default impl on subsequent calls because reading service file was skipped when jarread is true: XMLReader reader = XMLReaderFactory.createXMLReader(); System.out.println("1:" + reader.getClass().getName()); XMLReader reader2 = XMLReaderFactory.createXMLReader(); System.out.println("2:" + reader2.getClass().getName()); The fix is simply recover the original code. Here's the webrev: http://cr.openjdk.java.net/~joehw/jdk8/8016133/webrev/ Thanks, Joe

More information about the core-libs-dev mailing list