RFR(S): JDK-8044364: runtime/RedefineFinalizer test fails on windows
dmitry.samersoff at oracle.com
Sat May 31 10:50:32 UTC 2014
Thank you for the review!
Please, see also below.
On 2014-05-31 00:58, Daniel D. Daugherty wrote:
> On 5/30/14 1:59 PM, Dmitry Samersoff wrote:
>> Hi Everyone,
>> I rewrote test launcher in java.
>> Compilation left up to JTREG. So the test compiles with tools.jar from
>> COMPILEJAVA but run with tools.jar from TESTJAVA - the same way as all
>> other similar tests.
> I was expecting pages and pages of Java code. :-) Guess not.
I drop all compilation related staff - if COMPILEJAVA doesn't match
TESTJAVA, whole test run become messy anyway.
I also minimize error check - just check for "Redefinition done" to make
sure the test is valid.
> I'm fine with what you have here, but please make sure that
> you test it on the usual nightly platforms. We don't want
> another broken test in the nightly.
I had an impression that JPRT runs new JTREG tests before push, and it's
sad that I'was wrong here as I don't have all possible platforms
(including embedded) in hands.
> One question: any reason the test wasn't written using the
> <jdk-repo>/test/java/lang/instrument testing support and
> just use JLI's RedefineClass() function? Yes, I know that
> means the test would be in the <jdk-repo> instead of the HotSpot
> repo, but I don't know if anyone has added JLI testing
> infrastructure to the <hotspot-repo>.
I would prefer to use ASM to create instrumented class - we adopted it,
and I think it's a better way forward than home-grown ilib. Also I
prefer to keep the test in hotspot/test/runtime because this is the
tests against hotspot issue.
Oracle Java development team, Saint Petersburg, Russia
* I would love to change the world, but they won't give me the sources.
More information about the hotspot-runtime-dev