wasi:filesystem@0.3.0-rc-2025-09-16: Add tests for mkdir/rmdir #143
Conversation
wingo
force-pushed
the
filesystem-mkdir-rmdir
branch
2 times, most recently
from
995dfc0 to
836df64
Compare
wingo
changed the title
wingo
force-pushed
the
filesystem-mkdir-rmdir
branch
from
836df64 to
32d7e43
Compare
|
|
||
| async fn test_mkdir_rmdir(dir: &Descriptor) { | ||
| let mkdir = |path: &str| dir.create_directory_at(path.to_string()); | ||
| let rmdir = |path: &str| dir.remove_directory_at(path.to_string()); |
There was a problem hiding this comment.
Can you please add exploration of behavior further down the tree, where the test creates a dir and then opens it as a Descriptor and performs operations there, most interestingly the operations on .. and ../sibling?
There was a problem hiding this comment.
Hmm, just gave this a go @pchickey but I don't understand the behavior. Should one be able to work on .. from a child? For me it is failing. I thought that the capabilities were linked more to the set of directories exposed to WASI, and not to the individual directory objects.
There was a problem hiding this comment.
Well I certainly did not grasp this wrinkle of the WASI filesystem model, that one cannot .. from a subdir! To me it seems like nonsense to enforce a permission boundary between parts of a guest for files that are by definition reachable from the preopens. I can see why (because of the limitations of RESOLVE_BENEATH and the fact that subdirs can be moved around) but it still seems like nonsense to me. Filed WebAssembly/WASI#856, probably won't go anywhere but one can try.
wingo
force-pushed
the
filesystem-mkdir-rmdir
branch
2 times, most recently
from
632b27a to
e890570
Compare
wingo
force-pushed
the
filesystem-mkdir-rmdir
branch
from
e890570 to
8406bbf
Compare
wingo
force-pushed
the
filesystem-mkdir-rmdir
branch
5 times, most recently
from
bf69575 to
e2dae8a
Compare
|
|
||
| // remove-directory-at: async func(path: string) -> result<_, error-code>; | ||
| assert_eq!(rmdir("").await, Err(ErrorCode::NoEntry)); | ||
| assert_eq!(rmdir(".").await, Err(ErrorCode::Invalid)); |
There was a problem hiding this comment.
On Windows:
thread '<unnamed>' (1) panicked at src/bin/filesystem-mkdir-rmdir.rs:72:5:
assertion `left == right` failed
left: Err(ErrorCode { code: 0, name: "access", message: "Permission denied, similar to `EACCES` in POSIX." })
right: Err(ErrorCode { code: 11, name: "invalid", message: "Invalid argument, similar to `EINVAL` in POSIX." })
There was a problem hiding this comment.
Added an exception to allow Access. Discouraging, but let's see if we manage to make it to the end of the file.
wingo
force-pushed
the
filesystem-mkdir-rmdir
branch
from
e2dae8a to
6bd123f
Compare
|
Blocked on wasmtime Windows bug: bytecodealliance/wasmtime#12178. |
2 participants
No description provided.