Fix python tests

[yan12125]

WIP - create a PR to see if tests are correctly run

---

Thank you for submitting a PR!

Please delete this standard text once you've created your own description.

If you make changes to any of the code generators (src/idl_gen*) be sure to
build your project, as it will generate code based on the changes. If necessary
the code generation script can be directly run (scripts/generate_code.py),
requires Python3. This allows us to better see the effect of the PR.

If your PR includes C++ code, please adhere to the
Google C++ Style Guide,
and don't forget we try to support older compilers (e.g. VS2010, GCC 4.6.3),
so only some C++11 support is available.

For any C++ changes, please make sure to run sh scripts/clang-format-git.sh

Include other details as appropriate.

Thanks!

[dbaileychess]

Could you also update the CI builds?

[musicinmybrain]

I have not reviewed the contents of this PR or checked whether it fully solves the referenced bugs.

I have verified that, after rebasing this PR to the 23.5.9 release and applying it as a patch to the Fedora Linux flatbuffers package, the Python tests again pass on all relevant architectures: x86_64, aarch64, ppc64le, s390x.

Still, I’m holding off on updating the package in Fedora Rawhide from 23.3.3 until things settle out a bit—probably waiting for the next release. (Since flatbuffers offers no ABI stability guarantees., it’s never updated in stable Fedora branches anyway.)

[dbaileychess]

@maxburke Can you comment, as you introduce these changes.

[maxburke]

@maxburke Can you comment, as you introduce these changes.

I remember that we used dot-prefixed imports to sort out some issues where Python couldn't resolve some paths. I can't remember the exact reason why. If there are tests that pass when this is merged that don't otherwise then let's merge this change and I can sort out any of the failures I'm encountering at a later date.

[yan12125]

Thanks, I pushed another commit to fix CI. This should be ready for reviews now.

[yan12125]

The test failure seems unrelated:

external/boringssl/src/crypto/x509/t_x509.c:321:18: error: variable 'l' set but not used [-Werror,-Wunused-but-set-variable]

(https://buildkite.com/bazel/flatbuffers/builds/8508#018826d1-4342-4d50-a470-4095f006948c)

[dbaileychess]

Thanks!