I had previously `test_locale` as expected to fail, as it did indeed
fail on my system due to unimplemented functionality. As it happens, it
passes in CI because the locale settings used there (`C`, I believe)
just happen to format integers the same with "%d" as "%n".
I mistakenly un-marked it because I thought I misunderstood the problem.
Technically speaking, my system was misconfigured, leading me to disable
the test in the first place.
`test_locale` calls `locale.setlocale(locale.LC_ALL, '')`, which reads
the value of the `LANG` environment variable and uses that to look up
and reset all the locale settings. My system has `LANG=en_US.UTF-8`,
which is apparently not what this test was expecting. If `LANG` is unset
or set to `C`, the test passes, as it does in CI.
There are a substantial number of socket tests that are disabled due to
`bind(): bad family` errors. It seems like RustPython only supports a
small subset of the required connection families, so the failing tests
are broken for the same reasons.
The failing test was unsetting `PYTHONPATH`, but neglecting to unset
`RUSTPYTHONPATH`, which obviously was not significant for the original
CPython test. Including `RUSTPYTHONPATH` in the test fixes it.
This test will not work until the `__buffer__` and `__release_buffer__`
methods are implemented on the appropriate builtin types, which is
outside the current scope.
* pin openssl version for windows CI
* use cargo vcpkg
* install openssl with vcpkg
* use Swatinem/rust-cache right after dtolnay/rust-toolchain
* cargo install --target-dir=target cargo-vcpkg
---------
Co-authored-by: Dmitry Erlikh <d.erlikh@smartrecruiters.com>
Fix benchmarks by replacing rust-cpython with pyo3
- Replace `cpython` and `python3-sys` dependencies with `pyo3`.
- Add `html_report` feature to `criterion`, it will be required in a
future release.
- Remove `anyhow`. It was unused and cargo cleaned it up automatically.
- Use `compile` and `exec` from within the CPython context to
compile and execute benchmarks. There's probably more overhead to that than
the internal API had, but that overhead should be consistent per
benchmark.
The original hash algorithm just XOR'd all the hashes of the elements of
the set, which is problematic. The CPython algorithm is required to pass
the tests.
- Replace `PyFrozenSet::hash` with CPython's algorithm
- Remove unused `hash_iter_unorded` functions
- Add `frozenset` benchmark
- Enable tests
- Lower performance expectations on effectiveness test
- Adjust `slot::hash_wrapper` so that it doesn't rehash the computed
hash value in the process of converting PyInt to PyHash.
Python does not define `list().__bool__`, `dict().__bool__`, and
`str().__bool__`, and some tests were failing because they were
defined.
I could not find any references to them and deleting them doesn't
seem to break anything.
Update the actual benchmark harnesses.
Because the internal APIs previously used are no longer available, I
opted to use `compile` and `exec` from within the CPython context to
compile and execute code. There's probably more overhead to that than
the internal API had, but that overhead should be consistent per
benchmark.
If anyone cares about hyperoptimizing benchmarks then they can optimize
the harness as well.
The benchmarks have been broken since Python 3.10 deprecated the API
they were using to parse and execute CPython baselines. Since then
rust-cpython has been deprecated in favor of pyo3.
- Replace `cpython` and `python3-sys` with `pyo3`.
- Add `html_report` feature to `criterion`, it will be required in a
future release.
- Remove `anyhow`. It was unused and cargo cleaned it up automatically.
Kirill Podoprigora