chore: bump to wasi 0.3.0-rc-2026-02-09 #12557
Conversation
| impl stdin::Host for WasiCliCtxView<'_> {} | ||
|
|
||
| impl stdout::HostWithStore for WasiCli { | ||
| async fn write_via_stream<U>( |
There was a problem hiding this comment.
The primary changes in this rc were driven by: WebAssembly/WASI#870
ricochet
force-pushed
the
rev-0.3.0-rc-2026-02-09
branch
from
9f0a2f4 to
5bf9c47
Compare
| let file = get_file(store.get().table, &fd)?; | ||
| if !file.perms.contains(FilePerms::WRITE) { | ||
| return Ok(FutureReader::new(&mut store, async { | ||
| wasmtime::error::Ok(Err(ErrorCode::NotPermitted)) | ||
| })); | ||
| } |
There was a problem hiding this comment.
For this, append_via_stream below, and send with TCP, the ? here and the permission check all need to go into a created FutureReader. Here the call to get_file failing in ? will trigger a trap as opposed to an error code, but the file-not-writable case is handled just fine here. In TCP down below the two ? present there need to be funneled into a returned FutureReader, however.
There was a problem hiding this comment.
This is applicable to read_directory below too a bit. One thing I might recommend is using and_then or similar combinators to avoid using ? in the function bodies and that should make it a bit easier to see the flow of things.
Another option would be to remove the trappable option for these imports in the bindgen! configuration. That'll remove the wasmtime::Result layer of the return value and force us to implement this in a non-trapping fashion.
ricochet
force-pushed
the
rev-0.3.0-rc-2026-02-09
branch
from
5bf9c47 to
458b3ba
Compare
2 participants
No description provided.