RFR: 8350983: JShell LocalExecutionControl only needs stopCheck() on backward branches [v2]
Jan Lahoda
jlahoda at openjdk.org
Tue Apr 22 11:40:51 UTC 2025
On Mon, 3 Mar 2025 15:19:22 GMT, Archie Cobbs <acobbs at openjdk.org> wrote:
>> JShell's `LocalExecutionControl` ensures a running eval thread can be stopped by `JShell.stop()` by inserting an invocation of `stopCheck()` at every branch point in the bytecode of loaded classfiles.
>>
>> However, this same guarantee can be provided by invoking `stopCheck()` only at backward branches and not at forward branches. By doing this, the performance and size impact on the executing code can be reduced.
>>
>> Fortunately the new ClassFile API makes this kind of modification easy 👍
>
> Archie Cobbs has updated the pull request incrementally with one additional commit since the last revision:
>
> Niceify using pattern matching switch with guard per review suggestion.
Some comments inlined.
src/jdk.jshell/share/classes/jdk/jshell/execution/LocalExecutionControl.java line 99:
> 97: ClassTransform.transformingMethodBodies(
> 98: CodeTransform.ofStateful(() -> {
> 99: HashSet<Label> priorLabels = new HashSet<>();
Nit: could be `Set<Label> priorLabels = new HashSet<>();`.
src/jdk.jshell/share/classes/jdk/jshell/execution/LocalExecutionControl.java line 102:
> 100: return (builder, element) -> {
> 101: switch (element) {
> 102: case LabelTarget target -> priorLabels.add(target.label());
I think we usually indentat for `case`/`default`, right? (The old way of putting `case`/`default` at the same indentation level as was the switch is not really used in new code, or is it?)
-------------
PR Review: https://git.openjdk.org/jdk/pull/23850#pullrequestreview-2783802231
PR Review Comment: https://git.openjdk.org/jdk/pull/23850#discussion_r2053930429
PR Review Comment: https://git.openjdk.org/jdk/pull/23850#discussion_r2053932201
More information about the kulla-dev
mailing list