Signed-off-by: David Runge <dvzrv@archlinux.org>

Ignore AUTHORS file as names and mail addresses may contain unknown
words.
Ignore GIR as special term.

Signed-off-by: David Runge <dvzrv@archlinux.org>

Signed-off-by: David Runge <dvzrv@archlinux.org>

Python-six is a piece of legacy software that should be removed wherever
it is used.
As such we replace the import with importlib from Python.

Signed-off-by: David Runge <dvzrv@archlinux.org>

Suggested-by: Jelle van der Waa <jelle@vdwaa.nl>

Commit f4ef0624 changed `AbstractRule` so it inherits from `ABC`. This
inadvertently introduced a regression so the `all_rules` dict became
empty, essentially breaking all of namcap.

The existing logic in `__init__.py` that populates `all_rules` uses a
filter which tries to let only things pass that are classes:

    type(v) == type

However, this really means "check that v is a class with no superclass,"
so this check only allows top-level classes (inheriting directly from
`object`) to pass.

That still worked by accident because there was no inheritance at all,
but letting `AbstractRule` inherit from `ABC` finally broke it.
To fix this regression, change the check to:

    isinstance(v, type)

which means "check that v is a class."

There is no unit test that tests (or interacts with) `all_rules`, which
explains why this went past CI checks.

Only use comprehensions when the expression it contains is an actual
value. The `.append` function returns nothing useful, so replace those
two comprehensions with a classic for loop.

Note: the added assertion is not a test case assertion but just a
typing hint.

I would have used TestCase’s `assertIsNotNone` method, but that alone
wouldn’t convince mypy that the union type cannot be None:

> error: Item "None" of "Match[str] | None" has no attribute "group"  [union-attr]

These issues were caused by type annotations introduced in 4e0331f2.
In 96f36723, the fallout was partially fixed in `licensepkg.py` but not
the rest of the codebase.

Fix the issue for all occurrences I could spot.
I mainly inspected all `arg-type` violations, but I also fixed other
occurrences that mypy can’t detect yet whenever I bumped into one.

CC: David Runge <dvzrv@archlinux.org>

This basic mypy [1] configuration suppresses all type-checking errors
that it would currently report given the codebase as-is.

The idea is to re-enable those errors one by one and then fix them,
the ultimate goal being to make maintenance easier (due to editors
making more helpful suggestions and diagnosis) and uncover hidden bugs.

Besides, add a `files` directive so one can run `mypy` without
arguments.

Requires `extra/mypy` [2].

[1]: https://www.mypy-lang.org/
[2]: https://archlinux.org/packages/extra/any/mypy/

Pekka Ristola authored 1 year ago and Caleb Maclennan committed 1 year ago

With this change, `+` SPDX operator is allowed for all license
identifiers except those that end with `-only` or `-or-later`.
Syntactically the SPDX specification allows using the operator for any
license identifier, but semantically it doesn't make sense to use it for
licenses like `GPL-2.0-only+`. Such uses will be reported as unknown
license identifiers.

While there are other licenses that could probably be disallowed like
`MIT+`, it's hard to draw the line on which identifers should support
the `+` operator. The SPDX specification doesn't provide that
information, see https://github.com/spdx/spdx-spec/issues/689.

The functionality is implemented by removing trailing `+` from the
license symbols, since the python `license_expression` module doesn't
support parsing it directly. See
https://github.com/nexB/license-expression/issues/9.

Jelle van der Waa authored 1 year ago and Caleb Maclennan committed 1 year ago

Install the invalid.py "binary" so it is checked by namcap.

Jelle van der Waa authored 1 year ago and Caleb Maclennan committed 1 year ago

sphinx-build by default leaves unreproducible pickle files and
other a cache files which should not be packaged.

Jelle van der Waa authored 1 year ago and Caleb Maclennan committed 1 year ago

For PKGBUILD's with split packages the pkgbase does not have to be a
"valid" package or contain a pkgdesc.

Fixes #77

The `lstrip` function treats its argument as a set of characters
to remove in any order, while `removeprefix` removes a prefix.

Jelle van der Waa authored 1 year ago and Caleb Maclennan committed 1 year ago

Now CI should detect max line length violations.

Jelle van der Waa authored 1 year ago and Caleb Maclennan committed 1 year ago

- Fix listing rules to list all rules and not exit 2
- Allow listing rules and make packages not always mandatory
- Fix the parsing of rules/exclude they are comma separated
- Make --info a bool, as it is one
- Fix typo packag => package
- Format with ruff
- Remove unused import of dedent previously used for getopt

Co-authored-by: Caleb Maclennan <caleb@alerque.com>