Stream: git-wasmtime

Topic: wasmtime / issue #3291 Refactor the internals of `Store<T>`

Wasmtime GitHub notifications bot (Sep 03 2021 at 15:08):

alexcrichton commented on issue #3291:

It's worth pointing out that one refactoring I actually tried out was "what if we used Pin<Box<StoreInner<T>>>? I actually got pretty far with this but there's a few reasons that I didn't commit to it:

• Primarily, I don't think Pin is the right abstraction. My goal is to represent how when you have Pin<&mut StoreInner<T>> or Pin<&mut StoreOpaque> you have to be careful about accessing the contents and not overwrite the entire structure. I misunderstood the guarantees of Pin, though. The Pin::set method is safe and allows, at any time, overwriting the contents of the pinned pointer. That means that whatever gymnastics we go through to use Pin doesn't actually provide us any more safety since it's still possibly to safely cause bad behavior.

• Otherwise Pin was primarily pretty unergonomic. The main point against it was that, as we do today, we are required to manually reborrow Pin every time it's passed around and used twice.

So basically the ergonomic drop plus not actually providing much benefit led me to the conclusion that we simply need to continue to be vigilant within Wasmtime to do the right thing. We do still need the PhantomPinned, though, I think. I'm at least somewhat sure that we can't be marking all our pointers as noalias with all the aliasing we actually have...

Wasmtime GitHub notifications bot (Sep 03 2021 at 15:40):

github-actions[bot] commented on issue #3291:

Subscribe to Label Action

cc @peterhuene

<details>
This issue or pull request has been labeled: "wasmtime:api"

Thus the following users have been cc'd because of the following labels:

• peterhuene: wasmtime:api

To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.

Learn more.
</details>

Last updated: Nov 22 2024 at 17:03 UTC