Executors.newVirtualThreadPerTaskExecutor() leaks CompletableFuture

[Paul Bjorkstrand]

That is not unique to Executors.newVirtualThreadPerTaskExecutor. It is also
the same Executors.newThreadPerTaskExecutor. Both use the
same ThreadPerTaskExecutor (with different thread factories) that returns
instances of ThreadBoundFuture, which extends CompletableFuture [1], for
calls to the submit(..) methods.

I don't think that is "leaking" in the way a leak is typically considered
(i.e. a memory leak). It may be "leaking" some information about the
implementation, but that is just a consequence of an open, inspectable
system like the JDK.

[1]:
https://github.com/openjdk/loom/blob/fibers/src/java.base/share/classes/java/util/concurrent/ThreadPerTaskExecutor.java#L327

-Paul


On Tue, Mar 14, 2023 at 3:41 PM Kasper Nielsen <kasperni at gmail.com> wrote:

> Hi,
>
> I don't know if it is much of an issue in practice but you can cast the
> futures returned from Executors.newVirtualThreadPerTaskExecutor
> to CompletableFuture:
>
> ExecutorService e = Executors.newVirtualThreadPerTaskExecutor();
> CompletableFuture f = (CompletableFuture<?>) e.submit(() -> {});
>
> Other executors return instances of FutureTask which is probably less
> likely to be abused.
>
> /Kasper
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/loom-dev/attachments/20230314/407e9421/attachment.htm>