Repository navigation
Conversation
|
@DePasqualeOrg Is this CI actually testing wasip1 target as we expected even it is missing |
|
I don’t have the full context on this issue, and I don’t have the time to fully understand it, but I did spend about an hour working on this with Opus 5.5 in Claude Code, which produced the following concise explanation. I hope that’s helpful. Yes, it tests the wasip1 build. The binary is built for wasm32-wasip1, and the host-compiled tests run it through wasmtime ( One real gap: tests gated on |
|
But unix, not(target_os = "wasi"), etc... should NOT compile test for wasi. If we applied custom cfg for many tests, it is really noisy for our code base.
Cannot we fix it and use the most natural way?
|
|
I'm sorry, but I'm overloaded at the moment. I've spent another half hour working with Claude Code to resolve this, but I suggest you use these same tools to resolve any open questions, since it doesn't make sense for me to act as a proxy. Below is the analysis from Claude Code.
The noise could still be reduced: most #14886 no longer adds any custom cfg. It removes the dead |
|
Sorry. But I still don't agree with usefullness of wasi_runner. It looks a hack. @eduardomourar Is there any idea to remove |
|
What about moving these to a skip list in |
|
What do you mean? I DON'T agree with ignoreing tests using custom cfg: |
|
Claude Code (Opus 5.5) says |
For me, it is crazy. |
|
OK. By the way, do you still want to ignore them instead of compiling out? I think cfg(not(wasi_runner))) is better.
|
|
Yes, ignoring within the integration tests still gives us the better Developer Experience. Otherwise, we will have to maintain a separate script for the WASI tests (that would include the list of skipped tests). |
No description provided.