RFR: 8355323: JShell LocalExecutionControl should add stopCheck() at method entry
Chen Liang
liach at openjdk.org
Thu Apr 24 21:16:01 UTC 2025
On Thu, 24 Apr 2025 20:35:28 GMT, Archie Cobbs <acobbs at openjdk.org> wrote:
> This PR addresses [JDK-8355323](https://bugs.openjdk.org/browse/JDK-8355323), which inserts a "stop check" at the start of each method when JShell is running in local execution mode. This augments the existing stop checks that are added at each backward branch, which, while helpful, do not guarantee the snippet thread will notice a stop request in a bounded length of time, because a snippet could be executing some recursive method(s) that have no backward branches (an example is given in the new regression test). This fix closes that loophole.
>
> Side notes:
> * The existing `StopExecutionTest` unit test was split/refactored to create an `AbstractStopExecutionTest` superclass which could be reused by the new `LocalStopExecutionTest`.
> * There are three disabled tests in `StopExecutionTest` that seem to be working now and maybe could be re-enabled.
src/jdk.jshell/share/classes/jdk/jshell/execution/LocalExecutionControl.java line 102:
> 100: CodeTransform.ofStateful(() -> {
> 101: Set<Label> priorLabels = new HashSet<>();
> 102: AtomicBoolean methodEntry = new AtomicBoolean();
Instead of doing this hack, I think you should do so by overriding `CodeTransform.atStart` instead. Unfortunately we did not provide a utility for it like `endHandler`.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/24859#discussion_r2059239413
More information about the kulla-dev
mailing list