On Mon, 26 May 2025 17:08:45 GMT, Aleksey Shipilev <sh...@openjdk.org> wrote:
> SonarCloud complains that since > [JDK-8356080](https://bugs.openjdk.org/browse/JDK-8356080) we are allocating > and using `Boolean` boxes in `ReverseOrderListView`. This change `boolean` -> > `Boolean` was made in > [JDK-8356080](https://bugs.openjdk.org/browse/JDK-8356080) to allow `@Stable` > folding of boolean field. But it is very awkward to trade in the object > allocation on common path to allow optional constant folding. > > We can flatten this field to `byte` and check the specific non-zero values. > The field is final, so it is never actually in `0` state. > > Additional testing: > - [x] Linux x86_64 server fastdebug, `java/util` Well, this does at least make `ReverseOrderListView` have the same layout size as before [JDK‑8356080] and no longer grow when `‑XX:‑UseCompressedOops` is used. -------------------------------------------------------------------------------- Also, shouldn’t `ReverseOrderListView`’s `final` fields already be trusted to be immutable? [JDK‑8356080]: https://bugs.openjdk.org/browse/JDK-8356080 ------------- PR Comment: https://git.openjdk.org/jdk/pull/25456#issuecomment-2911387045
