Useless conversion error when using PyResult<PyObject> in pyo3

[edlng]

Summary

In rustc v1.84 we didn't have this issue but after upgrading to 1.85, we started getting this error. Downgrading back to 1.84 seemed to fix the issue.

Reproducer

For example I tried this dummy code:

  #[pyfunction]
  fn sum_as_string(py: Python) -> PyResult<PyObject> {
      let py_dict = PyDict::new_bound(py);

      Ok(py_dict.into_py(py))
  }

I expected to see no warnings or errors

Instead, this happened:

error: useless conversion to the same type: `pyo3::PyErr`
   --> src/lib.rs:160:45
    |
160 |     fn sum_as_string(py: Python) -> PyResult<PyObject> {
    |                                             ^ help: consider removing
    |
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#useless_conversion

Version

Additional Labels

No response

[LilyAcorn]

I'm hitting this too and I've used https://github.com/rust-lang/cargo-bisect-rustc to bisect to rust-lang/rust@19e75f4.

cargo bisect-rustc --start=1.84.0 --end=1.85.0 -c clippy -- clippy -- -F clippy::useless_conversion

I'm not immediately sure how to bisect further, but I'll spend a bit more time this evening to see if I can work it out.

[LilyAcorn]

I think I managed to bisect to 2ddfc92, but now it's passing and I don't know why...

[LilyAcorn]

Assuming that's the right commit, I now suspect this requires a fix in PyO3's #[pyfunction] and #[pymethods] proc macros, though I'm a little surprised we didn't pick up this before 1.85 was released.

[LilyAcorn]

Actually, this does appear to be fixed on v0.23.5 of PyO3 - I guess my code was using an older version. @edlng I suspect this is also the case for you.

[samueltardieu]

Assuming that's the right commit, I now suspect this requires a fix in PyO3's #[pyfunction] and #[pymethods] proc macros, though I'm a little surprised we didn't pick up this before 1.85 was released.

As the author of the commit you identified, which indeed lints more cases of useless conversions than it did before, I also had a look and could not reproduce it, and suspected a proc macro problem as well. I haven't looked at the pyfunction or pymethods proc macros, but sometimes macro writers use incorrect spans during expansion (for example through quote::quote_spanned!()) which prevent tools such as Clippy from noticing that the linted code comes from expansion of an external macro. This can lead to inappropriate diagnostics.

If you can exhibit a minimal compilable reproducer that I can use, I'll look at it, unless you consider the problem fixed already, in which case this issue can be closed.

[LilyAcorn]

I think there might be more to do in PyO3, but right now I don't think there's anything wrong in Clippy.

[edlng]

Thanks! I'll take a look again in the morning and close the issue once I confirm it is fixed.