This issue appears specifically for TypedDicts because the TypedDict constructor
code converts string annotations to ForwardRef objects, and those are not evaluated
properly by the get_type_hints() stack because of other shenanigans with type
parameters.
This issue does not affect normal generic classes because their annotations are not
pre-converted to ForwardRefs.
The fix attempts to restore the pre- #137227 behavior in the narrow scenario where
the issue manifests. It mostly makes changes only in the paths accessible from get_type_hints(),
ensuring that newer APIs (such as evaluate_forward_ref() and annotationlib) are not affected
by get_type_hints()'s past odd choices. This PR does not fix issue #138949, an older issue I
discovered while playing around with this one; we'll need a separate and perhaps more
invasive fix for that, but it should wait until after 3.14.0.
This amends commit bf8bbe9a81 by
restricting `echo_char` in `getpass.getpass` to single printable
ASCII characters as it would be uncommon to use long strings or
multi-byte characters for keyboard feedback.
---------
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Co-authored-by: Brian Schubert <brianm.schubert@gmail.com>
Rely on default test discovery.
Validation:
```bash
# Run before commit
./python -m test test_io -uall,walltime,largefile,cpu,extralargefile -M25G -o --fail-env-changed -j0 --list-cases | sort > old_cases.txt
# Run after commit
./python -m test test_io -uall,walltime,largefile,cpu,extralargefile -M25G -o --fail-env-changed -j0 --list-cases | sort > new_cases.txt
diff new_cases.txt old_cases.
# <outputs no changes in case list>
```
There was a deadlock originally seen by Memray when a daemon thread
enabled or disabled profiling while the interpreter was shutting down.
I think this could also happen with garbage collection, but I haven't
seen that in practice.
The daemon thread could be hung while trying acquire the global rwmutex
that prevents overlapping global and per-interpreter stop-the-world events.
Since it already held the main interpreter's stop-the-world lock, it
also deadlocked the main thread, which is trying to perform interpreter
finalization.
Swap the order of lock acquisition to prevent this deadlock.
Additionally, refactor `_PyParkingLot_Park` so that the global buckets
hashtable is left in a clean state if the thread is hung in
`PyEval_AcquireThread`.
The stack collector base class keeps all frames until export() is
called, which causes significant unnecessary memory usage. Instead, we
can process the frames on the fly in the collect call by dispatching the
aggregation logic to the subclass through the process_frames method.
Co-authored-by: Pablo Galindo Salgado <pablogsal@gmail.com>
Rename `pathlib._os.magic_open()` to `vfsopen()`. The new name is a bit
less abstract, and it aligns with the `vfspath()` method added in 5dbd27d.
Per discussion on discourse[^1], adjust `vfsopen()` so that the following
methods may be called:
- `__open_reader__()`
- `__open_writer__(mode)`
- `__open_updater__(mode)`
These three methods return readable, writable, and full duplex file objects
respectively. In the 'writer' method, *mode* is either 'a', 'w' or 'x'. In
the 'updater' method, *mode* is either 'r' or 'w'.
In the pathlib ABCs, replace `ReadablePath.__open_rb__()` with
`__open_reader__()`, and replace `WritablePath.__open_wb__()` with
`__open_writer__()`.
[^1]: https://discuss.python.org/t/open-able-objects/90238
Co-authored-by: Petr Viktorin <encukou@gmail.com>
The codecs lookup function now performs only minimal normalization of
the encoding name before passing it to the search functions:
all ASCII letters are converted to lower case, spaces are replaced
with hyphens.
Excessive normalization broke third-party codecs providers, like
python-iconv.
Revert "bpo-37751: Fix codecs.lookup() normalization (GH-15092)"
This reverts commit 20f59fe1f7.
While file timestamps can be anything the file system can store, most
lie between the recent past and the near future. Optimize fill_time()
for typical timestamps in three ways:
- When possible, convert to nanoseconds with C arithmetic.
- When using C arithmetic and the seconds member is not required (for
st_birthtime), avoid creating a long object.
- When using C arithmetic, reorder the code to avoid the null checks
implied in Py_XDECREF().
Co-authored-by: Victor Stinner <vstinner@python.org>
Adds a --no-randomize option to the CI runner, so that randomisation can be easily
disabled for --fast-ci and --slow-ci configurations on single-threaded testing platforms
like Android, iOS, and Emscripten.
---------
Co-authored-by: Malcolm Smith <smith@chaquo.com>
This makes information about the interpreter ABI more accessible.
Co-authored-by: Petr Viktorin <encukou@gmail.com>
Co-authored-by: Victor Stinner <vstinner@python.org>
Co-authored-by: Adam Turner <9087854+AA-Turner@users.noreply.github.com>
* KeyError is not raised for defaultdict
* Fix relative paths on different drives on Windows
* Add a round-trip test
Co-authored-by: Itamar Oren <itamarost@gmail.com>
The signature algorithms allowed for certificate-based client authentication or
for the server to complete the TLS handshake can be defined on a SSL context via
`ctx.set_client_sigalgs()` and `ctx.set_server_sigalgs()`.
With OpenSSL 3.4 or later, the list of available TLS algorithms can be retrieved
by `ssl.get_sigalgs()`.
With OpenSSL 3.5 or later, the selected signature algorithms can be retrieved from
SSL sockets via `socket.client_sigalg()` and `socket.server_sigalg()`.
This commit also partially amends 377b787618
by using `PyUnicode_DecodeFSDefault` instead of `PyUnicode_DecodeASCII` in
`_ssl._SSLContext.get_groups`, so that functions consistently decode strings
obtained from OpenSSL.
---------
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
