7173494: some jdk tests are not run in test/Makefile (original) (raw)

Alan Bateman Alan.Bateman at oracle.com
Mon Oct 8 18:12:33 UTC 2012

On 08/10/2012 18:44, Mandy Chung wrote:

It's good to clean this up and the change looks fine in general. A couple of minor comments:

jdkmanagement - Might be good to include java/lang/management tests in this target in case someone only runs one target to verify that area. On the other hands, they are currently covered by jdklang target and might be better to take them out from jdklang (maybe in the future). The jdk_management target could run the java/lang/management tests too but it would mean these tests would run twice. I have experimented with moving these tests to test/java/lang_management (same thing for java/lang/instrument to java/lang_instrument) but decided not to include those changes here as it's a disruptive change. With compact profiles [1] and future modularization then we may need to look at this again as it would be really nice to be able to specify the test directories to jtreg that correspond to the modules or profiles (I think it's just a bit early to know how the tests will run). jdkrmi target - you change "javax/rmi" to "javax/rmi/ssl". There is no test in the test/javax/rmi directory currently but if a new test (or new subdirectory) is added under test/javax/rmi, you would have to change the jdkrmi target. Would it be better to keep "javax/rmi" as it is? javax.rmi.CORBA has the classes for RMI-IIOP and I would expect tests for this area to be run by a different make target, say jdk_corba. As it happens we don't have any tests for this area in the jdk repository so it doesn't make any difference at this time So somewhat of a drive-by fix (again profiles and modules motivated so that "make jdk_rmi" would run the RMI tests only).

test/sun/misc seems to belong to jdklang. Do you know why they are in jdkother? The tests in test/sun/misc provide very sparse coverage and are somewhat of a mixed bag so I decided to leave them where they are. I wouldn't object to moving them to the jdk_lang target, do you have a strong opinion on this?

-Alan

[1] http://openjdk.java.net/jeps/161

More information about the core-libs-dev mailing list