[jdk11u-dev] RFR: 8231454: File lock in Windows on a loaded jar due to a leak in Introspector::getBeanInfo

[Basil Crow]

On Tue, 24 May 2022 04:04:12 GMT, Basil Crow <duke at openjdk.java.net> wrote:

> Hi all,
> 
> This pull request contains a backport of [JDK-8231454](https://bugs.openjdk.java.net/browse/JDK-8231454) (openjdk/jdk at 2ee2b4ae19d).
> 
> There was a trivial merge conflict in `Introspector.java`.
> 
> The commit being backported was authored by @mrserb on October 20, 2020, and was reviewed by @azuev-java.
> 
> See [JENKINS-63766](https://issues.jenkins.io/browse/JENKINS-63766) for the motivation behind this backport.
> 
> I tested this PR locally by running `make run-test TEST=test/jdk/java/beans/Introspector/FlushClassInfoCache.java`. In addition, I verified that JENKINS-63766 could no longer be reproduced with this backport applied.
> 
> Thanks!

I do not have JBS access. Can someone please submit the following comment with the `jdk11u-fix-request` label:

> See [JENKINS-63766](https://issues.jenkins.io/browse/JENKINS-63766) for the motivation behind this backport. Jenkins users complained about a metaspace leak, and I was able to see the issue by running with `-verbose:class` on OpenJDK 9-15 but not on OpenJDK 8, 16, or 17. Bisection showed this issue was resolved in openjdk/jdk at 2ee2b4ae19d, and I verified locally in an end-to-end testing scenario that backporting openjdk/jdk at 2ee2b4ae19d to `jdk11u-dev` resolves the leak I was experiencing in JENKINS-63766 on OpenJDK 11. I also successfully executed the test from openjdk/jdk at 2ee2b4ae19d locally by running `make run-test TEST=test/jdk/java/beans/Introspector/FlushClassInfoCache.java`. The change applied cleanly other than a trivial merge conflict in `Introspector.java`.

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

PR: https://git.openjdk.java.net/jdk11u-dev/pull/1103