Gyeongjae Choi

Thanks for the report. I can reproduce the error, and the error happens in the [`FinalizationRegistry`](https://github.com/pyodide/pyodide/blob/4fbbbedc09496c6968086d69aadba75398718b13/src/core/pyproxy.ts#L82), which handles garbage collecting the proxy objects. Looks like we are trying to decref...

As you commented in the original PR (and as sbc100 suggested), maybe we can stop using reportUndefinedSymbols after landing RPATH changes. Also, reportUndefinedSymbols was no op for 1 month after...

Closing as resolved

The latest version of `@types/emscripten` version is 1.40, but I believe they are not targeting Emscripten 1.40 (which was released in 5 years ago), so we should probably check its...

@R3gardless Assigned to you. Thanks! FYI, this issue mentions only the `FS` object, but improving the type definition using the `@types/emscripten` can be one of the long-term goals in https://github.com/pyodide/pyodide/issues/5786...

Yes, Pyodide uses some private emscripten APIs internally, so they should be separately defined as a type for internal usage. But for the exported FS module, we do not want...

We recently adopted DefinitelyTyped/Emscripten for FS type, thanks to @R3gardless. So the IDEs should be able to use types from DefinitelyTyped/Emscripten.

It looks like skipping (host) numpy installation is not working after this change, resulting in scipy build error.

I am having the statsmodels build issue in [pyodide-recipes PR](https://github.com/pyodide/pyodide-recipes/pull/72) as well. It uses tot pyodide-build, so I think it is not related to pyodide/pyodide-build#60. Maybe there was other regression.

Yes, I am planning to provide a wheelhouse using Anaconda package index: - https://anaconda.org/pyodide/ - https://anaconda.org/pyodide-build/ I am publishing wheels built in [pyodide-recipes](https://github.com/pyodide/pyodide-recipes) (which, I hope, will replace in-tree recipes...