task: quickly send task to heap on debug mode

[blasrodri]

Ref: #2055

Motivation

Solution

[Darksonn]

That's certainly one way to approach it. I have some ideas for alternatives, but I'm ok with merging this for now, possibly replacing it with another idea later if it works out.

Please note that we have multiple spawn methods. It would also be an idea to have a test that spawns a massive future to catch regressions in how large futures we can handle.

[blasrodri]

Good idea! Do you want to include the regression test in this PR?

[Darksonn]

Sure.

[blasrodri]

That's certainly one way to approach it. I have some ideas for alternatives, but I'm ok with merging this for now, possibly replacing it with another idea later if it works out.

Please note that we have multiple spawn methods. It would also be an idea to have a test that spawns a massive future to catch regressions in how large futures we can handle.

Is there a way to get the limit value of the stack? I assume that the one running in the CI/CD is different than the one on my local machine. I can test a few, but perhaps there was a better way.

[Darksonn]

The original issue had a link to a function called remaining_stack, but I don't think you can change the size of the future dynamically based on its value. A better alternative might be to go through the thread builder and set your own stack size directly.

[hawkw]

Alternatively, couldn't we just make the majority of the functions the task is passed through be #[inline(always)]?

[Darksonn]

According to this comment, that apparently didn't help as much as you would hope.

[hawkw]

@Darksonn huh, that's weird! never mind, then.

[blasrodri]

The original issue had a link to a function called remaining_stack, but I don't think you can change the size of the future dynamically based on its value. A better alternative might be to go through the thread builder and set your own stack size directly.

I'm trying to get back to this one, so it can be merged.
Your proposal is to set a fix stack through the builder, and then create a task that's equally big, right? I guess we also need to create a test set on the CI that sets the opt-level in order to reproduce the debug setup?