Read a negative address value from the stack after switching to rustc 1.85

[tomasol]

Describe the Bug

I have a WASM project that is using prost-wkt-types. After switching rustc from 1.84 to 1.85 wasm-bindgen started failing with:

thread 'main' panicked at /build/wasm-bindgen-cli-0.2.100-vendor/wasm-bindgen-wasm-interpreter-0.2.100/src/lib.rs:245:21:
prost_wkt_types::pbstruct::_::_::__ctor::he32f530e01fca88f: Read a negative address value from the stack. Did we run out of memory?
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: wasm_bindgen_wasm_interpreter::Interpreter::call
   3: wasm_bindgen_wasm_interpreter::Interpreter::call
   4: wasm_bindgen_wasm_interpreter::Interpreter::call
   5: wasm_bindgen_wasm_interpreter::Interpreter::interpret_descriptor
   6: wasm_bindgen_cli_support::descriptors::execute
   7: wasm_bindgen_cli_support::Bindgen::generate_output
   8: wasm_bindgen_cli_support::Bindgen::generate
   9: wasm_bindgen::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

Steps to Reproduce

I have created a reproducer repo. The issue is visible as the GitHub Action result or locally as described in the readme.

Expected Behavior

It works with rustc 1.84

Actual Behavior

Fails with rustc 1.85

Additional Context

Not sure where the actual error comes from rustc, wasm-bindgen, or prost-wkt-types

[laskinner]

I'm having this issue as well in Leptos. App compiles and runs fine with 1.84, but 1.85 and above throw the same error.