Tips on handling `ffi-check`?

[clin1234]

In my fork, when running CI tests, against Python 3.14, errors spew in the form of these:

error[E0609]: no field `ob_base` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_weakreflist` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_name` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_qualname` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_exc_state` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_origin_or_finalizer` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_hooks_inited` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_closed` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_running_async` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_frame_state` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `gi_iframe` on type `_PyGenObject`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

error[E0609]: no field `_padding` on type `_object`
  --> src/main.rs:45:13
   |
45 |             pyo3_ffi_check_macro::for_all_fields!($name, check_field);
   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
...
70 |     pyo3_ffi_check_macro::for_all_structs!(check_struct);
   |     ---------------------------------------------------- in this macro invocation
   |
   = note: this error originates in the macro `pyo3_ffi_check_macro::for_all_fields` which comes from the expansion of the macro `pyo3_ffi_check_macro::for_all_structs` (in Nightly builds, run with -Z macro-backtrace for more info)

Reproducible by running cargo run --manifest-path=pyo3-ffi-check/Cargo.toml

Any tips derived from adding prior CPython releases would be appreciated.

[clin1234]

For context, compare the definition of PyGenObject in 3.14 vs in 3.13

[ngoldbaum]

The way to fix this is to add support for 3.14 in pyo3-ffi and update the FFI definitions.

We generally don't add support for new Python versions until the first beta, since the API and ABI aren't stable until then.

That said, because of #4904 I think there's a case to be made to do that a little early this year so we can add TSAN testing on the free-threaded build using 3.14, which has a number of thread safety fixes that haven't been backported to 3.13.

[ngoldbaum]

I think I'm going to open and maintain a draft PR people can use to test Python 3.14 against. That way we can quickly merge the PR when 3.14 beta 1 comes out.

[clin1234]

I think I'm going to open and maintain a draft PR people can use to test Python 3.14 against. That way we can quickly merge the PR when 3.14 beta 1 comes out.

I already have a PR: #4811

[clin1234]

Albeit still needs fixing FFI...

[ngoldbaum]

Ah missed you'd been pushing on that. Do you mind if I push to your PR?

[clin1234]

Sure, though my PR adds 3.14 for testing in CI, and the FFI test is a vexing pain in the collective posterior of all pyo3 developers...

[ngoldbaum]

Let's close this since the issues with ffi-check have been resolved in the PR. I'm happy to share more about how I was working with it to debug ffi issues, but I was more or less just reading the error messages and going from there to what was wrong in the FFI.