gh-105699: Use a _Py_hashtable_t for the PyModuleDef Cache (gh-106974)
This fixes a crasher due to a race condition, triggered infrequently when two isolated (own GIL) subinterpreters simultaneously initialize their sys or builtins modules. The crash happened due the combination of the "detached" thread state we were using and the "last holder" logic we use for the GIL. It turns out it's tricky to use the same thread state for different threads. Who could have guessed?
We solve the problem by eliminating the one object we were still sharing between interpreters. We replace it with a low-level hashtable, using the "raw" allocator to avoid tying it to the main interpreter.
We also remove the accommodations for "detached" thread states, which were a dubious idea to start with.
(cherry picked from commit 8ba4df91ae)
The _xxsubinterpreters module was meant to only use public API. Some internal C-API usage snuck in over the last few years (e.g. gh-28969). This fixes that.
(cherry picked from commit e6373c0d8b)
Co-authored-by: Eric Snow <ericsnowcurrently@gmail.com>
gh-105059: Fix MSCV compiler warning on PyObject union (GH-107239)
Use pragma to ignore the MSCV compiler warning on the PyObject
nameless union.
(cherry picked from commit 1c8fe9bdb6)
Co-authored-by: Victor Stinner <vstinner@python.org>
gh-105059: Use GCC/clang extension for PyObject union (GH-107232)
Anonymous union is new in C11. To prevent compiler warning
when using -pedantic compiler option, use Clang and GCC
extension on C99 and older.
(cherry picked from commit 6261585d63)
Co-authored-by: Victor Stinner <vstinner@python.org>
gh-98608: Move PyInterpreterConfig to pylifecycle.h (GH-107191)
Move PyInterpreterConfig structure and associated macros from
initconfig.h to pylifecycle.h: it's not related to the Python
Initialization Configuration.
(cherry picked from commit e717b47ed8)
Co-authored-by: Victor Stinner <vstinner@python.org>
GH-96803: Move PyUnstable_InterpreterFrame_GetCode() to Python.h (GH-107188)
Declare the following 3 PyUnstable functions in
Include/cpython/pyframe.h rather than Include/cpython/frameobject.h,
so they are now provided by the standard "GH-include <Python.h>".
(cherry picked from commit 837fa5c0cd)
Co-authored-by: Victor Stinner <vstinner@python.org>
GH-103082: Rename PY_MONITORING_EVENTS to _PY_MONITORING_EVENTS (#107069)
Rename private C API constants:
* Rename PY_MONITORING_UNGROUPED_EVENTS to _PY_MONITORING_UNGROUPED_EVENTS
* Rename PY_MONITORING_EVENTS to _PY_MONITORING_EVENTS
(cherry picked from commit 0927a2b25c)
gh-105227: Add PyType_GetDict() (GH-105747)
This compensates for static builtin types having `tp_dict` set to `NULL`.
(cherry picked from commit a840806d33)
Co-authored-by: Eric Snow <ericsnowcurrently@gmail.com>
Co-authored-by: Petr Viktorin <encukou@gmail.com>
gh-105340: include hidden fast-locals in locals() (GH-105715)
* gh-105340: include hidden fast-locals in locals()
(cherry picked from commit 104d7b760f)
Co-authored-by: Carl Meyer <carl@oddbird.net>
gh-106140: Reorder some fields to facilitate out-of-process inspection (GH-106143)
(cherry picked from commit 2d5a1c2811)
Signed-off-by: Pablo Galindo <pablogsal@gmail.com>
Co-authored-by: Pablo Galindo Salgado <Pablogsal@gmail.com>
gh-105587: Remove assertion from `_PyStaticObject_CheckRefcnt` (GH-105638)
(cherry picked from commit 6199fe3b32)
Co-authored-by: Eddie Elizondo <eduardo.elizondorueda@gmail.com>
gh-105387: Limited C API implements Py_INCREF() as func (GH-105388)
In the limited C API version 3.12, Py_INCREF() and Py_DECREF()
functions are now implemented as opaque function calls to hide
implementation details.
(cherry picked from commit b542972dc1)
Co-authored-by: Victor Stinner <vstinner@python.org>
For a while now, pending calls only run in the main thread (in the main interpreter). This PR changes things to allow any thread run a pending call, unless the pending call was explicitly added for the main thread to run.
(cherry picked from commit 757b402)
We are changing it to be more flexible that a strict bool can be for possible future expanded used cases.
(cherry picked from commit b97e14a806)
Co-authored-by: Eric Snow <ericsnowcurrently@gmail.com>
* gh-102304: Fix Py_INCREF() for limited C API 3.9 (#105550)
When Python is built in debug mode (Py_REF_DEBUG macro), Py_INCREF()
and Py_DECREF() of the limited C API 3.9 (and older) now call
Py_IncRef() and Py_DecRef(), since _Py_IncRef() and _Py_DecRef() were
added to Python 3.10.
(cherry picked from commit 7ba0fd9f87)
* gh-102304: Remove Py_INCREF() doc change (#105552)
Py_INCREF() was made compatible again with Python 3.9 and older in
the limited API of Python debug mode.
(cherry picked from commit 58e4b69f69)
The risk of a race with this state is relatively low, but we play it safe anyway. We do avoid using the lock in performance-sensitive cases where the risk of a race is very, very low.
(cherry picked from commit 68dfa49627)
Co-authored-by: Eric Snow <ericsnowcurrently@gmail.com>
The risk of a race with this state is relatively low, but we play it safe anyway.
(cherry picked from commit 7799c8e678)
Co-authored-by: Eric Snow <ericsnowcurrently@gmail.com>
gh-102304: Fix Py_INCREF() stable ABI in debug mode (#104763)
When Python is built in debug mode (if the Py_REF_DEBUG macro is
defined), the Py_INCREF() and Py_DECREF() function are now always
implemented as opaque functions to avoid leaking implementation
details like the "_Py_RefTotal" variable or the
_Py_DecRefTotal_DO_NOT_USE_THIS() function.
* Remove _Py_IncRefTotal_DO_NOT_USE_THIS() and
_Py_DecRefTotal_DO_NOT_USE_THIS() from the stable ABI.
* Remove _Py_NegativeRefcount() from limited C API.
(cherry picked from commit 92022d8416)
This avoids the problematic race in drop_gil() by skipping the FORCE_SWITCHING code there for finalizing threads.
(The idea for this approach came out of discussions with @markshannon.)
(cherry picked from commit 3698fda)
Co-authored-by: Eric Snow ericsnowcurrently@gmail.com
In gh-103912 we added tp_bases and tp_mro to each PyInterpreterState.types.builtins entry. However, doing so ignored the fact that both PyTypeObject fields are public API, and not documented as internal (as opposed to tp_subclasses). We address that here by reverting back to shared objects, making them immortal in the process.
(cherry picked from commit 7be667d)
Co-authored-by: Eric Snow ericsnowcurrently@gmail.com
This commit replaces the Python implementation of the tokenize module with an implementation
that reuses the real C tokenizer via a private extension module. The tokenize module now implements
a compatibility layer that transforms tokens from the C tokenizer into Python tokenize tokens for backward
compatibility.
As the C tokenizer does not emit some tokens that the Python tokenizer provides (such as comments and non-semantic newlines), a new special mode has been added to the C tokenizer mode that currently is only used via
the extension module that exposes it to the Python layer. This new mode forces the C tokenizer to emit these new extra tokens and add the appropriate metadata that is needed to match the old Python implementation.
Co-authored-by: Pablo Galindo <pablogsal@gmail.com>
This PR updates `math.nextafter` to add a new `steps` argument. The behaviour is as though `math.nextafter` had been called `steps` times in succession.
---------
Co-authored-by: Mark Dickinson <mdickinson@enthought.com>
