On Tue, 11 Aug 2026 06:25:13 GMT, Sorna Sarathi N <[email protected]> wrote:

>> The test was using a fixed -XX:MaxMetaspaceSize=25m for all platforms. 
>> Investigation (via jcmd VM.metaspace) revealed that the higher limit is 
>> needed only on platforms where CDS is unavailable (e.g. AIX). Without CDS, 
>> all 3,340 classes are allocated in classic metaspace (~23.4 MB committed), 
>> whereas on Linux with CDS active, 1,364 core JDK classes are mapped from the 
>> shared archive into a separate "Shared Class Space" that does not count 
>> against -XX:MaxMetaspaceSize, keeping classic metaspace usage at ~13.6 MB.
>> 
>> This was validated by running the test on Linux with -Xshare:off 
>> -XX:MaxMetaspaceSize=17m, which reproduces the same OOM as AIX. When running 
>> with 25m, the usage is very similar to AIX.
>> 
>> This change splits the test into two variants based on CDS availability:
>> 
>>> id=cds (@requires vm.cds): runs with -XX:MaxMetaspaceSize=17m
>> id=nocds (@requires !vm.cds): runs with -XX:MaxMetaspaceSize=25m
>> 
>> An explanatory comment is also included so the reasoning is clear to future 
>> readers.
>> 
>> JBS Issue: [JDK-8336666](https://bugs.openjdk.org/browse/JDK-8336666)
>> 
>> ---------
>> - [x] I confirm that I make this contribution in accordance with the 
>> [OpenJDK Interim AI Policy](https://openjdk.org/legal/ai).
>
> Sorna Sarathi N has updated the pull request incrementally with one 
> additional commit since the last revision:
> 
>   Minor Fix

test/hotspot/jtreg/serviceability/jvmti/RedefineClasses/RedefineLeakThrowable.java
 line 54:

> 52:  */
> 53: 
> 54: // This test has two variants with different MaxMetaspaceSize values 
> depending on CDS availability.

also maybe move this comment on top ?

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/32254#discussion_r3755693719

Reply via email to