RFR: 8300032: DwarfParser resource leak
Chris Plummer
cjplummer at openjdk.org
Thu Jan 12 21:29:15 UTC 2023
On Thu, 12 Jan 2023 12:08:51 GMT, Daniel Jeliński <djelinski at openjdk.org> wrote:
> Please review this fix for DwarfParser cleaner.
>
> The original code registered the cleaner using a lambda that captured a reference to the parser object; as a result, the object was never GCed, and the cleaner never ran.
>
> In this version I moved the lambda creation to a static method, so that it can't capture a reference to the parser.
>
> Additionally, the new code uses a static cleaner; the cleaner object creation and cleanup are heavy operations (every cleaner comes with its own thread), and it's preferable to use a single shared cleaner where possible.
>
> I verified manually that the native `destroyDwarfContext` method is called after this fix. No new automated test. Existing tier1-2 tests continue to pass.
This seems like a pretty serious flaw with using Lambdas and Cleaners that probably should be brought up with the libs and language teams. I see other occurrences of this issue in these files (I may have missed some):
src/java.base/share/classes/com/sun/crypto/provider/*Key.java (4 files)
src/java.base/windows/classes/java/lang/ProcessImpl.java
src/java.desktop/share/classes/javax/swing/border/TitledBorder.java
src/java.desktop/share/classes/javax/swing/filechooser/FileSystemView.java
-------------
PR: https://git.openjdk.org/jdk/pull/11965
More information about the serviceability-dev
mailing list