Hi again,
Could you please review an update?
Turns out that the ResourceCheck test didn't work for Mac when I removed
it from the problems list. There was code in the shell script that needed
to know all the platform names.
Instead of just adding "Darwin" to the script I decided to make the test
pure
Java, to speed things up and to avoid similar platform problems in the
future. I couldn't find a way in JTREG to add a test-JDK relative
library using
tags so I loaded the lib/jconsole.jar dynamically within the test.
Changes from previous webrev:
- classes will be loaded using a URLClassLoader (test.jdk/lib/jconsole.jar)
- the resource bundle is loaded from the custom class loader
- the method Resources#getMnemonicInt is now invoked by reflection. Not
super clean, but better then having a shell script in my opinion.
- removed failing tests from the ProblemsList.txt
Tested in JPRT with "-testset core -onlytests '.*jdk_tools2.*'" on all
platforms,
Here is the webrev:
http://cr.openjdk.java.net/~egahlin/7172176_02/
Thanks for taking the time to review this.
Erik
Erik Gahlin skrev 2012-05-30 13:40:
Hi,
Could you please review 7172176, which is a test fix for 7017818 -
JConsoleResources.java cannot be handled by translation team.
Webrev:
http://cr.openjdk.java.net/~egahlin/7172176_01/
Changes:
- Removed the ImmutableRequest.java and ImmutableRequest.sh since the
JConsoleResources class was refactored away with 7017818.
- Removed hard-wired resource bundle keys in ResourceCheckTest.java,
the keys
are now looked up by reflection on the Messages class. The test also
checks there is one-to-one mapping between reource bundle keys and
the constants available in the Message class.
Michael,
I removed sun/tools/jconsole/ResourceCheckTest.sh from the jdk/test/
ProblemList.txt. This means the test will break once the translated files
will be checked in. Unused messages were sent for translation and they
need to be removed.
Are you ok with that?
Thanks!
Erik
