Matthew Davidson

@arnaudgeiser I'm not clear how your linked code sample relates. What I'm contemplating is this: We alter the code in `chain-/chain'-` to stop invoking the callback immediately in `(let [x''...

Zach, thanks for chiming in! You're always welcome here. I think the proposed change of checking the executor and continuing to run in the current thread if it's already in...

Maybe I need to take a closer look, but I think the general point is, with our own pools, we can alter Dirigiste/Manifold/Aleph if needed to achieve this. To the...

Zach, IIUC, you're suggesting the second thread local contain the executor the thread is _actually_ running in, and that it will never change (Which is unlike how the `instrumented-executor`'s `:onto?`...

FYI, bit busy at the moment, I won't be able to get back to this until the following week.

Probably, I just haven't done it yet because it's a bit tedious :) There's 40+ methods in the CompletionStage interface that would have to be implemented, probably for multiple deferred...

That would certainly be one way to consolidate code, but at the cost of an extra layer of indirection for all Manifold deferred calls. I don't think that tradeoff is...

FWIW, Zach wrote some funky macro attempt at inheritance that might help in this situation, if you can get it working. Checkout Potemkin's `def-abstract-type` and `deftype+` macros. It's not truly...

Adding a potemkin dep is fine, especially since Aleph already depends on it, but double-check, because there are a few macros that _were_ copied from potemkin, and don't have the...

No, we don't have to care as much about speed in tests, so unrolling isn't needed. However, maybe consider the alternative of copying those JDK tests over? Their test methods...