RFR: 8286681: ShenandoahControlThread::request_gc misses the case of GCCause::_codecache_GC_threshold
David Holmes
dholmes at openjdk.java.net
Fri May 13 13:22:03 UTC 2022
On Fri, 13 May 2022 13:02:52 GMT, Jie Fu <jiefu at openjdk.org> wrote:
>> I assume you are running the tests with:
>> make run-tests TEST_OPTS_JAVA_OPTIONS="-XX:+UseShenandoahGC"
>> in which case, all of the tests you select to run will be run with that GC.
>>
>> What you have is not wrong but wouldn't be a better to add a new test to test/hotspot/jtreg/gc rather than relying on test in test/jdk/java/foreign?
>
>> I think I agree with @AlanBateman - in the sense that this seems to go down a slippery slope where every test would need to be executed against all possible GCs. AFAIK, there are no other foreign tests doing this.
>
> I think it's a waste of time to write a separate test for this bug.
>
> I can't understand why you are against adding one more test config in the foreign test.
> Yes, it caught the bug in ShenandoahGC but who knows it wouldn't find some other bugs in the foreign api in the future.
> If you are still against the newly added test config, I can revert the test change.
> Thanks.
My comment seems to have never made it through. The problem with what your are doing is two-fold:
1. When running the test suite specifically to test xGC for whatever x you are now forcing a test run for Shenandoah.
2. When x is Shenandoah then you run this test twice.
You don't need a config just to run this with Shenandoah - you run the test suite with Shenandoah.
-------------
PR: https://git.openjdk.java.net/jdk/pull/8691
More information about the core-libs-dev
mailing list