RFR: 8282170: JVMTI SetBreakpoint metaspace allocation test
Aleksey Shipilev
shade at openjdk.java.net
Tue Mar 1 08:22:05 UTC 2022
On Tue, 22 Feb 2022 19:54:59 GMT, Chris Plummer <cjplummer at openjdk.org> wrote:
>> There are few bugs in SetBreakpoint when it reaches for metaspace allocation, notably [JDK-8214992](https://bugs.openjdk.java.net/browse/JDK-8214992) and [JDK-8264149](https://bugs.openjdk.java.net/browse/JDK-8264149). This adds the test that exercises the metaspace allocation paths.
>>
>> Requires [JDK-8282172](https://bugs.openjdk.java.net/browse/JDK-8282172) to pass cleanly in fastdebug mode.
>>
>> Additional testing:
>> - [x] New test on Linux x86_64 fastdebug
>> - [x] New test on Linux x86_64 release
>> - [x] New test with [JDK-8264149](https://bugs.openjdk.java.net/browse/JDK-8264149) reverted fails on Linux x86_64 release
>
> So you load a class 50 times in separate ClassLoaders, and each load triggers the ClassPrepare event, and each time that happens you set a breakpoint on each of the 1000 methods, meaning a total of 5000 breakpoints. Doing this will eventually trigger calling CollectorPolicy::satisfy_failed_metadata_allocation(). Previous to the [8264149](https://bugs.openjdk.java.net/browse/JDK-8264149) fix, this used to trigger an assert. Am I understanding this properly?
Thanks @plummercj! Anyone else to review?
-------------
PR: https://git.openjdk.java.net/jdk/pull/7554
More information about the serviceability-dev
mailing list