fitzgen (he/him) (Jun 12 2020 at 17:45): fitzgen (he/him) (Jun 12 2020 at 17:45):implementing Func::wrap support for reference types right now, and WasmTy::{load,store} need to take a &Store argument to pass in/receive ExternRefs to/from Wasm, which means that the trampoline in impl_into_func! needs to get a Store from ... somewhere? presumably from either the callee_vmctx or caller_vmctx?
cc @Alex Crichton
Alex Crichton (Jun 12 2020 at 17:46):hm... so the only guaranteed non-null value is caller_vmctx (the first one)
fitzgen (he/him) (Jun 12 2020 at 17:46):ah, host_state()?
Alex Crichton (Jun 12 2020 at 17:46):nah that probably doesn't have a Store
Alex Crichton (Jun 12 2020 at 17:46):er or it might
fitzgen (he/him) (Jun 12 2020 at 17:47):just above trampoline, in shim:
debug_assert!(state.is::<(F, Weak<StoreInner>)>());
ah yeah no nvmd looks like it's there
fitzgen (he/him) (Jun 12 2020 at 17:47):thanks
Alex Crichton (Jun 12 2020 at 17:47):so yeah that looks to be it
Alex Crichton (Jun 12 2020 at 17:47):but I'd avoid upgrading the Weak unelss an ExternRef is actually used
Alex Crichton (Jun 12 2020 at 17:48):either way is fine though
fitzgen (he/him) (Jun 12 2020 at 17:48):good call
fitzgen (he/him) (Jun 12 2020 at 17:51):ah hm, that's actually a little funky -- are you thinking that WasmTy::{load,store} should instead take a &dyn Fn() -> &Store or a &Weak<StoreInner>? because we don't know at the call site what the T: WasmTy is and whether it will use the store or not
fitzgen (he/him) (Jun 12 2020 at 17:51):@Alex Crichton
Alex Crichton (Jun 12 2020 at 17:51):either way I think is fine
Alex Crichton (Jun 12 2020 at 17:51):ExternRef stores a weak anyway, right?
Alex Crichton (Jun 12 2020 at 17:52):it's an internal trait FWIW (the mehtods at least) so it's fine to do whatever works best
fitzgen (he/him) (Jun 12 2020 at 17:52):ah yeah, so I guess it is only WasmTy::load that really needs it
Last updated: Dec 06 2025 at 06:05 UTC