Part of #7928

python/cpython@fdc0e09

Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>

Co-authored-by: hauntsaninja <>

Co-authored-by: hauntsaninja <>

…ction (#7920)

Co-authored-by: Sebastian Rittau <srittau@rittau.biz>
Co-authored-by: Akuli <akuviljanen17@gmail.com>

python/cpython@c379bc5

#7906 (comment)
I felt this better documents how seed is used in practice

Co-authored-by: hauntsaninja <>

Required for #7865

python/cpython@35b98e3

* Annotations for psycopg2.ConnectionInfo

These annotations come from the documentation here:

https://www.psycopg.org/docs/extensions.html#psycopg2.extensions.ConnectionInfo
If there was doubt, I referred to the libpq documentation cited by
psycopg2's docs.

I wasn't completely sure about `dsn_parameters`. Psycopg2's docs list it
as an `dict`, and the example suggests it's a `dict[str, str]` at that.
From psycopg2's source I found

    https://github.com/psycopg/psycopg2/blob/1d3a89a0bba621dc1cc9b32db6d241bd2da85ad1/psycopg/conninfo_type.c#L183-L206

which is implemented here:

    https://github.com/psycopg/psycopg2/blob/1d3a89a0bba621dc1cc9b32db6d241bd2da85ad1/psycopg/utils.c#L251-L279

I'm no expert in CPython's API, but this looks to me like it's building
a `dict[str, str]`.

Additionally, the libpq docs

https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-PQCONNINFO
https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-PQCONNDEFAULTS

show that the underlying data just consists of strings.

Additionally, I'm pretty sure from this chunk of source

    https://github.com/psycopg/psycopg2/blob/1d3a89a0bba621dc1cc9b32db6d241bd2da85ad1/psycopg/conninfo_type.c#L581-L598

That `ConnectionInfo.__init__` takes one positional-only argument, which
must be a `psycopg2.connection`. But I don't think users are intended to
be constructing this type, so I've not added that annotation.

* Annotate `connection.info` and related attributes

* Make ConnectionInfo attributes properties

According to https://github.com/psycopg/psycopg2/blob/1d3a89a0bba621dc1cc9b32db6d241bd2da85ad1/psycopg/conninfo_type.c#L534-L563

* Mark connection attributes as readonly

according to https://github.com/psycopg/psycopg2/blob/8ef195f2ff187454cc709d7857235676bb4176ee/psycopg/connection_type.c#L1244

* Explain why some properties aren't `T | None`

- New attributes in 3.11
- Arguments to CodeType.__init__ are always positional (have fun)
- co_consts can use object instead of Any

Unblocks PyCQA/flake8-pyi#222

* zipfile: 3.11 fixes

Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>

The script doesn't import tomli

…ic (#7897)

* Don't run stubtest_stdlib on PRs that don't touch the stdlib

Similar errors are already ignored in Python 3 stubs. The
errors break Python 2 tests of mypy.

See #7367 (comment)
for context.

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.