packaging is the standard tool for dealing with versions and
requirements. This allows us to easily support all PEP 508 specs.
packaging also has good error messages, so I don't create our own here.
This lets us get rid of our handrolled half parser and sanity checks.
I kept the whitespace consistency check, since that's just a formatting
lint and isn't testing the correctness of parsing (which we now
definitely do correctly).
Currently, when a maintainer opens a pull request from a branch in the
typeshed repository (instead of a fork), CI is run twice: once for the
PR and once for the push. This not only wastes CI minutes (and it seems
we're running into some limits lately), but also makes CI output twice
as large and more difficult to process.
The disadvantage is that a PR needs to be opened to run CI.
Alternatively, CI can be triggered manually.
stubtest_third_party.py will now only print a single line when the test
succeeds for a certain distribution, unless the "-v" option is given.
When the test fails or "-v" is given, will still print the full output.
The success status is now colored to make spotting failures easier.
stdout/stderr is now used consistently: The distribution name and
success status is always printed to stdout, all other output goes to
stderr. Running the script with "2>/dev/null" will only show the success
status of the test, one per line.
Only changes to files in `stdlib` or `stubs` will ever have any effect on the checked open-source code. But also run primer on changes to files in `.github/workflows` so that we can see the effect of changes to the `mypy_primer.yml` file itself.
- mypy_test and pyright no longer just test the stubs, they now also test other parts of typeshed as well.
- pytype_test.py can now be run on 3.10, meaning the whole test suite can now also be run on 3.10.
- Various test scripts now have from `__future__ import annotations`, meaning they can now only be run on 3.7+.
- Clean up the description of pyright_test.py, which had a slightly confusing wording.
- Also fix the `--dry-run` config option in mypy_test.py, which I accidentally broke in #7746
* Upgrade pyright to 1.1.230
* Add `type: ignore`s for new pyright checks regarding multiple inheritance
* Temporarily pin pyright-action to 1.0.4, as changes made in 1.0.5 break typeshed's CI
Co-authored-by: Jelle Zijlstra <jelle.zijlstra@gmail.com>
* Run mypy on the 3.11 stdlib in CI, as a regression test for https://github.com/python/mypy/issues/12220
* Correct the docstring at the top of the file following the changes made in #7478
* Stop the test from crashing when run locally if there's an extra file/folder in the stubs directory.
Calling this "stubtest.yml" is a little confusing, since most of the
time stubtest is actually run out of "tests.yml". I renamed the workflow
in #7306 and this gets the filename to match.
Co-authored-by: hauntsaninja <>
Fixes#7344 for now.
As discussed in that issue, I think it might make sense to revert the
change to check positional-only arguments on dunders for now, since it
seems to be mostly busywork and has very little benefit for end users.
However, that'll happen on stubtest master, which typeshed isn't yet
ready for.
Co-authored-by: hauntsaninja <>
