gh-117549: Don't use designated initializers in headers (#118580)
The designated initializer syntax in static inline functions in pycore_backoff.h
causes problems for C++ or MSVC users who aren't yet using C++20.
While internal, pycore_backoff.h is included (indirectly, via pycore_code.h)
by some key 3rd party software that does so for speed.
Victor Stinner [Sun, 5 May 2024 10:15:19 +0000 (12:15 +0200)]
gh-110850: Remove _PyTime_TimeUnchecked() function (#118552)
Use the new public Raw functions:
* _PyTime_PerfCounterUnchecked() with PyTime_PerfCounterRaw()
* _PyTime_TimeUnchecked() with PyTime_TimeRaw()
* _PyTime_MonotonicUnchecked() with PyTime_MonotonicRaw()
gh-118164: Break a loop between _pydecimal and _pylong and optimize int to str conversion (GH-118483)
For converting large ints to strings, CPython invokes a function in _pylong.py,
which uses the decimal module to implement an asymptotically waaaaay
sub-quadratic algorithm. But if the C decimal module isn't available, CPython
uses _pydecimal.py instead. Which in turn frequently does str(int). If the int
is very large, _pylong ends up doing the work, which in turn asks decimal to do
"big" arithmetic, which in turn calls str(big_int), which in turn ... it can
become infinite mutual recursion.
This change introduces a different int->str function that doesn't use decimal.
It's asymptotically worse, "Karatsuba time" instead of quadratic time, so
still a huge improvement. _pylong switches to that when the C decimal isn't
available. It is also used for not too large integers (less than 450_000 bits),
where it is faster (up to 2 times for 30_000 bits) than the asymptotically
better implementation that uses the C decimal.
Eric Snow [Sat, 4 May 2024 21:24:02 +0000 (15:24 -0600)]
gh-117953: Track Extra Details in Global Extensions Cache (gh-118532)
We have only been tracking each module's PyModuleDef. However, there are some problems with that. For example, in some cases we load single-phase init extension modules from def->m_base.m_init or def->m_base.m_copy, but if multiple modules share a def then we can end up with unexpected behavior.
With this change, we track the following:
* PyModuleDef (same as before)
* for some modules, its init function or a copy of its __dict__, but specific to that module
* whether it is a builtin/core module or a "dynamic" extension
* the interpreter (ID) that owns the cached __dict__ (only if cached)
This also makes it easier to remember the module's kind (e.g. single-phase init) and if loading it previously failed, which I'm doing separately.
Victor Stinner [Sat, 4 May 2024 07:39:58 +0000 (09:39 +0200)]
gh-110850: Use _PyDeadline_Get() in EnterNonRecursiveMutex() (#118556)
Use _PyDeadline_Init() and _PyDeadline_Get() in
EnterNonRecursiveMutex() of thread_nt.h.
_PyDeadline_Get() uses the monotonic clock which is now the same as
the perf counter clock on all platforms. So this change does not
cause any behavior change. It just reuses existing helper functions.
Sam Gross [Fri, 3 May 2024 22:16:45 +0000 (18:16 -0400)]
gh-118527: Intern filename, name, and qualname in code objects. (#118558)
This interns the strings for `co_filename`, `co_name`, and `co_qualname`
on codeobjects in the free-threaded build. This partially addresses a
reference counting bottleneck when creating closures concurrently. The
closures take the name and qualified name from the code object.
Barney Gale [Fri, 3 May 2024 20:29:25 +0000 (21:29 +0100)]
GH-116380: Move pathlib globbing implementation into `pathlib._glob` (#118562)
Moving this code under the `pathlib` package makes it quite a lot easier
to backport in the `pathlib-abc` PyPI package. It was a bit foolish of me
to add it to `glob` in the first place.
Also add `translate()` to `__all__` in `glob`. This function is new in
3.13, so there's no NEWS needed.
Brett Simmers [Fri, 3 May 2024 15:30:55 +0000 (08:30 -0700)]
gh-116322: Add Py_mod_gil module slot (#116882)
This PR adds the ability to enable the GIL if it was disabled at
interpreter startup, and modifies the multi-phase module initialization
path to enable the GIL when loading a module, unless that module's spec
includes a slot indicating it can run safely without the GIL.
PEP 703 called the constant for the slot `Py_mod_gil_not_used`; I went
with `Py_MOD_GIL_NOT_USED` for consistency with gh-104148.
A warning will be issued up to once per interpreter for the first
GIL-using module that is loaded. If `-v` is given, a shorter message
will be printed to stderr every time a GIL-using module is loaded
(including the first one that issues a warning).
mpage [Fri, 3 May 2024 15:14:26 +0000 (08:14 -0700)]
gh-118495: Skip test using threads after forking when running with TSAN (#118530)
This is unsupported. Note that `skip_unless_reliable_fork()` checks for
the conditions used by the decorators that were removed, along with checking
for TSAN.
Sam Gross [Fri, 3 May 2024 15:09:57 +0000 (11:09 -0400)]
gh-117514: Add `sys._is_gil_enabled()` function (#118514)
The function returns `True` or `False` depending on whether the GIL is
currently enabled. In the default build, it always returns `True`
because the GIL is always enabled.
Sam Gross [Fri, 3 May 2024 15:05:30 +0000 (11:05 -0400)]
gh-118527: Use `_Py_ID(__main__)` for main module name (#118528)
Most module names are interned and immortalized, but the main
module was not. This partially addresses a scaling bottleneck in the
free-threaded when creating closure concurrently in the main module.
Brett Simmers [Thu, 2 May 2024 22:25:36 +0000 (15:25 -0700)]
gh-116738: Make `_codecs` module thread-safe (#117530)
The module itself is a thin wrapper around calls to functions in
`Python/codecs.c`, so that's where the meaningful changes happened:
- Move codecs-related state that lives on `PyInterpreterState` to a
struct declared in `pycore_codecs.h`.
- In free-threaded builds, add a mutex to `codecs_state` to synchronize
operations on `search_path`. Because `search_path_mutex` is used as a
normal mutex and not a critical section, we must be extremely careful
with operations called while holding it.
- The codec registry is explicitly initialized as part of
`_PyUnicode_InitEncodings` to simplify thread-safety.
Sam Gross [Thu, 2 May 2024 17:41:15 +0000 (13:41 -0400)]
gh-118413: Fix test_release_task_refs on free-threaded build (#118494)
The `time.sleep()` call should happen before the GC to give the worker
threads time to clean-up their remaining references to objs.
Additionally, use `support.gc_collect()` instead of `gc.collect()`
just in case the extra GC calls matter.
Mark Shannon [Thu, 2 May 2024 15:17:59 +0000 (16:17 +0100)]
GH-118095: Unify the behavior of tier 2 FOR_ITER branch micro-ops (GH-118420)
* Target _FOR_ITER_TIER_TWO at POP_TOP following the matching END_FOR
* Modify _GUARD_NOT_EXHAUSTED_RANGE, _GUARD_NOT_EXHAUSTED_LIST and _GUARD_NOT_EXHAUSTED_TUPLE so that they also target the POP_TOP following the matching END_FOR
gh-82062: Fix support of parameter defaults on methods in extension modules (GH-115270)
Now inspect.signature() supports references to the module globals in
parameter defaults on methods in extension modules. Previously it was
only supported in functions. The workaround was to specify the fully
qualified name, including the module name.
This commit uses "file pointed to by" to replace "file pointed by" in
- doc for shutil.copytree
- docstring for shutil.copytree
- docstring _abc.PathBase.open
- docstring for pathlib.Path.open
- doc for os.copy_file_range
- doc for os.splice
The docs use "file pointed to by" more frequently than
"file pointed by". So, this commit replaces the uses of
"file pointed by" in order to make the uses consistent
through the docs.
```bash
$ grep -ri 'pointed to by' cpython/
```
yields more results than
```bash
$ grep -ri 'pointed by' cpython/
```
Separately:
There are two occurrences of "tree pointed by":
- cpython/Doc/library/xml.etree.elementtree.rst for
`xml.etree.ElementInclude.include`
- cpython/Lib/xml/etree/ElementInclude.py for `include`
For those uses of "tree pointed by", I expect "tree pointed to by"
instead. However, I found enough uses online of (a) "tree pointed by"
rather than (b) "tree pointed to by" to convince me that (a) is in
common use.
So, this commit does not replace those occurrences of "tree pointed by"
to "tree pointed to by". But I will replace them if a reviewer
believes it is correct to replace them.
* docs: typo: "exists and executable" -> "exists and is executable"
mpage [Wed, 1 May 2024 20:59:12 +0000 (13:59 -0700)]
gh-118433: Temporarily skip `test_interrupt_main_subthread` in free-threaded builds (#118485)
Free-threaded builds can intermittently tickle a longstanding bug (24 years!)
in the implementation of `threading.Condition`, leading to flakiness in the
test suite. Fixing the underlying issue will require more discussion, and will
likely apply to most of the concurrency primitives in the `threading` module
that are written in Python. See gh-118433 for more details.
* Add documentation and tests. Tests release the GIL while calling
raw clock functions.
* py_get_system_clock() and py_get_monotonic_clock() now check that
the GIL is hold by the caller if raise_exc is non-zero.
* Reimplement "Unchecked" functions with raw clock functions.