Stream: git-wasmtime

Topic: wasmtime / issue #11222 Improve ergonomics and type-safet...

Wasmtime GitHub notifications bot (Jul 11 2025 at 17:54):

dicej opened issue #11222:

Currently, the implementation of {Future,Stream}Any and the conversion functions from those types to/from Host{Future,Stream} have a couple of major paper cuts:

• They're prone to leakage, e.g. dropping a StreamAny without closing it or passing it to the guest will leave the read end open, meaning any further writes will hang forever.
• Conversion to Host{Stream,Future} does no typechecking, and a type error will only be raised later when a read is matched with a write (and in some cases, even then there won't be a type error if the payload types are memcpy-able, e.g. u8 vs. s8).

Wasmtime GitHub notifications bot (Jul 11 2025 at 17:54):

dicej added the wasm-proposal:component-model-async label to Issue #11222.

Wasmtime GitHub notifications bot (Aug 12 2025 at 21:07):

dicej closed issue #11222:

Currently, the implementation of {Future,Stream}Any and the conversion functions from those types to/from Host{Future,Stream} have a couple of major paper cuts:

• They're prone to leakage, e.g. dropping a StreamAny without closing it or passing it to the guest will leave the read end open, meaning any further writes will hang forever.
• Conversion to Host{Stream,Future} does no typechecking, and a type error will only be raised later when a read is matched with a write (and in some cases, even then there won't be a type error if the payload types are memcpy-able, e.g. u8 vs. s8).

Wasmtime GitHub notifications bot (Aug 12 2025 at 21:07):

dicej commented on issue #11222:

@alexcrichton This seems to be a duplicate of #11161. I'm going to close this one in favor of that one; feel free to reopen if they're both needed.

Wasmtime GitHub notifications bot (Aug 12 2025 at 21:07):

dicej reopened issue #11222:

Currently, the implementation of {Future,Stream}Any and the conversion functions from those types to/from Host{Future,Stream} have a couple of major paper cuts:

• They're prone to leakage, e.g. dropping a StreamAny without closing it or passing it to the guest will leave the read end open, meaning any further writes will hang forever.
• Conversion to Host{Stream,Future} does no typechecking, and a type error will only be raised later when a read is matched with a write (and in some cases, even then there won't be a type error if the payload types are memcpy-able, e.g. u8 vs. s8).

Wasmtime GitHub notifications bot (Aug 12 2025 at 21:07):

dicej closed issue #11222:

Currently, the implementation of {Future,Stream}Any and the conversion functions from those types to/from Host{Future,Stream} have a couple of major paper cuts:

• They're prone to leakage, e.g. dropping a StreamAny without closing it or passing it to the guest will leave the read end open, meaning any further writes will hang forever.
• Conversion to Host{Stream,Future} does no typechecking, and a type error will only be raised later when a read is matched with a write (and in some cases, even then there won't be a type error if the payload types are memcpy-able, e.g. u8 vs. s8).

Last updated: Dec 06 2025 at 06:05 UTC