Support mayInterruptIfRunning flag in CompletableFuture.await extension

[elizarov]

This is useful for cases when we are working with interruptible blocking APIs. See discussion in #79

[qwwdfsad]

It cannot be implemented: CompletableFuture doesn't use flag mayInterruptIfRunning and classic Future is not a CompletionStage and is not awaitable

[qwwdfsad]

Citing https://docs.oracle.com/en/java/javase/17/docs/api/java.base/java/util/concurrent/CompletableFuture.html here:

mayInterruptIfRunning - this value has no effect in this implementation because 
interrupts are not used to control processing.

[revintec]

@qwwdfsad thanks for the link and I should have provided more details:

1. quoting your doc, this value has no effect in this implementation -- emphasis mine
   I don't think we should design APIs based (only) on current implementations. as this method is non-final, subclass may very well using this flag for finer controls

2. I did find a subclass in a very common code path -- awaiting HttpClient's async exchanges
   a CompletableFuture of type MinimalFuture returned from https://github.com/openjdk/jdk17/blob/74007890bb9a3fa3a65683a3f480e399f2b1a0b6/src/java.net.http/share/classes/jdk/internal/net/http/HttpClientImpl.java#L653

the corresponding coroutine waits on the resulting future using httpClient.sendAsync(...).await(), and subsequently get cancelled(due to timeout, etc). but the underlying http request can't be cancelled unless the mayInterruptIfRunning flag is correctly passed through -- as value of true, see
https://github.com/openjdk/jdk17/blob/74007890bb9a3fa3a65683a3f480e399f2b1a0b6/src/java.net.http/share/classes/jdk/internal/net/http/common/MinimalFuture.java#L107
https://github.com/openjdk/jdk17/blob/74007890bb9a3fa3a65683a3f480e399f2b1a0b6/src/java.net.http/share/classes/jdk/internal/net/http/MultiExchange.java#L276

[qwwdfsad]

@revintec Thanks!

Indeed there probably exist some implementations that use the flag, but this fact alone is not enough to expand our API surface. For now, it's not even clear whether these implementations from HttpClient leak to user code at all.

What we need is actual use cases where it is clear which exact user-facing problem we are solving with the proposed API addition

[dkhalanskyjb]

I'd think true would be a more idiomatic value to use here, so maybe we should also consider the alternative of just always passing true. When we cancel in coroutines, we almost always intend to cancel the computation if it was already started.

[fvasco]

Some hours ago I used HttpClient and looked for an awaitInterruptible method, similar to runInterruptible.

[dkhalanskyjb]

The problem is, is there anything in the HttpClient documentation that indicates that cancelling the future will interrupt the process? https://docs.oracle.com/en/java/javase/12/docs/api/java.net.http/java/net/http/HttpClient.html I don't see anything in that vein. So, should one rely on this behavior?

[fvasco]

The problem is, is there anything in the HttpClient documentation that indicates that cancelling the future will interrupt the process?

It works on JRE 16

https://bugs.openjdk.org/browse/JDK-8245462?focusedCommentId=14365751&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-14365751

[fvasco]

@dkhalanskyjb is it now possible to reopen this issue?