JDK-8060068/JDK-8067904 DriverManager clean-up introduces bootstrap problem that breaks production drivers

[Robert Gibson]

Hi again,
I just realised that I missed an important bit from the test case in my earlier mail.
The jar file used for the test needs to contain, in addition to the class files, a text file called META-INF/services/java.sql.Driver containing the text
MyDriver
since it is the interaction between the driver being loaded manually and through the Service Provider mechanism which causes the dysfunction.
Regards,
Robert

On Friday, 3 April 2015, 14:07, Robert Gibson <robbiexgibson at yahoo.com> wrote:



Thanks Alan,
Note that the repro given in my mail is a bit simpler than the current defect description, especially if you don't have access to the "MyFace" JDBC driver.
Regards,
Robert 




On Friday, 3 April 2015, 13:04, Alan Bateman <Alan.Bateman at oracle.com> wrote:
On 03/04/2015 10:33, Robert Gibson wrote:
> Hi there,
>
> We are doing some early testing with JDK 9 and have discovered that the changes made to java.sql.DriverManager back in November/December have introduced an incompatibility with our JDBC driver (that we have used unchanged since Java 6) when it is pre-loaded with Class.forName (as recommended by Tomcat and specified as being harmless in the JavaDoc). We filed a bug report but it got an incident ID JI-9019539 and then disappeared into a black hole.
>
There does appear to be an issue with those changes. I've located the 
incident and moved it to the JDK project as JDK-8076639 [1].

-Alan




[1] https://bugs.openjdk.java.net/browse/JDK-8076639