On Thu, 10 Sep 2026 15:11:13 GMT, Jorn Vernee <[email protected]> wrote:
> Implement a method handle combinator that can be used to synchronize on an
> object monitor while executing a given target method handle.
>
> The returned method handle behaves similar to the notional code:
>
>
> R adapter(Object lock, A... a) throws Throwable {
> synchronized (lock) {
> return body.invokeExact(a...);
> }
> }
>
>
> The lambda form is similar to existing ones: we box up all the arguments for
> the body handle, call a fallback function which synchronizes and invokes the
> body with the boxed argument, then the `Object` result is fed to an unboxing
> handle to unbox the result if needed.
>
> There's a corresponding intrinsic in `InvokerBytecodeGenerator` which
> replaces these three operations with `monitorEnter`/`monitorExit`
> instructions, direct argument/return value forwarding without (un)boxing, and
> the necessary exception handling which makes sure we unlock the lock again in
> case the body throws an exception. The generated code is similar to what
> javac generates.
>
> ---------
> - [x] I confirm that I make this contribution in accordance with the [OpenJDK
> Interim AI Policy](https://openjdk.org/legal/ai).
Don't know what prompted you to added this now (maybe FFM?) but this looks like
a reasonable addition in general. Some comments but these problems exist in
existing code like tableswitch forms.
src/java.base/share/classes/java/lang/invoke/MethodHandleImpl.java line 2272:
> 2270: }
> 2271:
> 2272: private static LambdaForm makeSynchronizeForm(MethodType basicType,
> BoundMethodHandle.SpeciesData data) {
The `data` is always BMH_LLL and you are already hardcoding the fields using
the `fieldCursor` below. I think you should drop this argument because it just
provides a false illusion of flexibility. (Though arguably the same problem
exists for the tableswitch form you added before)
src/java.base/share/classes/java/lang/invoke/MethodHandleImpl.java line 2326:
> 2324: }
> 2325:
> 2326: lform = LambdaForm.create(basicType.parameterCount() + 1,
> names, Kind.SYNCHRONIZE);
Side remark: we should really replace all `basicType.parameterCount() + 1` with
`ARG_LIMIT`.
-------------
PR Review: https://git.openjdk.org/jdk/pull/32817#pullrequestreview-5189478648
PR Review Comment: https://git.openjdk.org/jdk/pull/32817#discussion_r3998668470
PR Review Comment: https://git.openjdk.org/jdk/pull/32817#discussion_r3998686811
