This partially reverts #137047, keeping the tests for GC collectability of the
original class that dataclass adds `__slots__` to.
The reference leaks solved there are instead solved by having the `__dict__` &
`__weakref__` descriptors not tied to (and referencing) their class.
Instead, they're shared between all classes that need them (within
an interpreter).
The `__objclass__` ol the descriptors is set to `object`, since these
descriptors work with *any* object. (The appropriate checks were already
made in the get/set code, so the `__objclass__` check was redundant.)
The repr of these descriptors (and any others whose `__objclass__` is `object`)
now doesn't mention the objclass.
This change required adjustment of introspection code that checks
`__objclass__` to determine an object's “own” (i.e. not inherited) `__dict__`.
Third-party code that does similar introspection of the internals will also
need adjusting.
Co-authored-by: Jelle Zijlstra <jelle.zijlstra@gmail.com>
There were a few thread-safety issues when profiling or tracing all
threads via PyEval_SetProfileAllThreads or PyEval_SetTraceAllThreads:
* The loop over thread states could crash if a thread exits concurrently
(in both the free threading and default build)
* The modification of `c_profilefunc` and `c_tracefunc` wasn't
thread-safe on the free threading build.
Remove the `__dict__` and `__weakref__` descriptors from the original class when creating a dataclass from it.
An interesting hack, but more localized in scope than gh-135230.
This may be a breaking change if people intentionally keep the original class around
when using `@dataclass(slots=True)`, and then use `__dict__` or `__weakref__` on the
original class.
Co-authored-by: Alyssa Coghlan <ncoghlan@gmail.com>
Co-authored-by: Petr Viktorin <encukou@gmail.com>
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
The `PyEval_SetProfileAllThreads` function and other related functions
had a race condition on `tstate->c_profilefunc` that could lead to a
crash when disable profiling or tracing on all threads while another
thread is starting to profile or trace a a call.
There are still potential crashes when threads exit concurrently with
profiling or tracing be enabled/disabled across all threads.
Fix a bug caused by the garbage collector clearing weakrefs too early. The
weakrefs in the ``tp_subclasses`` dictionary are needed in order to correctly
invalidate type caches (for example, by calling ``PyType_Modified()``).
Clearing weakrefs before calling finalizers causes the caches to not be
correctly invalidated. That can cause crashes since the caches can refer to
invalid objects. Defer the clearing of weakrefs without callbacks until after
finalizers are executed.
In GH-116206, the comment about moving reachable objects to next generation
got moved from its original place to a place where there is no code below
it. Put the comment back to where the actual movement of reachable objects
happens.
Fix name of the Python encoding in Unicode errors of the code page
codec: use "cp65000" and "cp65001" instead of "CP_UTF7" and "CP_UTF8"
which are not valid Python code names.
Default implementation of sys.unraisablehook() now uses traceback._print_exception_bltin() to print exceptions with colorized text.
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Co-authored-by: Victor Stinner <vstinner@python.org>
Call backtrace() once when installing the signal handler to ensure that
libgcc is dynamically loaded outside the signal handler.
This fixes a "signal-unsafe call inside of a signal" TSan error from
test_faulthandler.test_enable_fd.
De-instrumenting code objects modifies the thread local bytecode for all threads as such, holding the critical section on the code object is not sufficient and leads to data races. Now, the de-instrumentation is now performed under a stop the world pause as such no thread races with executing the thread local bytecode while it is being de-instrumented.
Basic support for pyrepl in Emscripten. Limitations:
* requires JSPI
* no signal handling implemented
As followup work, it would be nice to implement a webworker variant
for when JSPI is not available and proper signal handling.
Because it requires JSPI, it doesn't work in Safari. Firefox requires
setting an experimental flag. All the Chromiums have full support since
May. Until we make it work without JSPI, let's keep the original web_example
around.
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
Co-authored-by: Éric <merwok@netwok.org>
This makes the following APIs public:
* `Py_BEGIN_CRITICAL_SECTION_MUTEX(mutex),`
* `Py_BEGIN_CRITICAL_SECTION2_MUTEX(mutex1, mutex2)`
* `void PyCriticalSection_BeginMutex(PyCriticalSection *c, PyMutex *mutex)`
* `void PyCriticalSection2_BeginMutex(PyCriticalSection2 *c, PyMutex *mutex1, PyMutex *mutex2)`
The macros are identical to the corresponding `Py_BEGIN_CRITICAL_SECTION` and
`Py_BEGIN_CRITICAL_SECTION2` macros (e.g., they include braces), but they
accept a `PyMutex` instead of an object.
The new macros are still paired with the existing END macros
(`Py_END_CRITICAL_SECTION`, `Py_END_CRITICAL_SECTION2`).
In free-threading, multiple threads can be cleared concurrently as such the modifications on `sys_tracing_threads` should be done while holding the profile lock, otherwise it can race with other threads setting up profiling.
`_datetime` is a special module, because it's the only non-builtin C extension that contains static types. As such, it would initialize static types in the module's execution function, which can run concurrently. Since static type initialization is not thread-safe, this caused crashes. This fixes it by moving the initialization of `_datetime`'s static types to interpreter startup (where all other static types are initialized), which is already properly protected through other locks.
This is useful for implementing proper `input()`. It requires the
JavaScript engine to support the wasm JSPI spec which is now stage 4.
It is supported on Chrome since version 137 and on Firefox and node
behind a flag.
We override the `__wasi_fd_read()` syscall with our own variant that
checks for a readAsync operation. If it has it, we use our own async
variant of `fd_read()`, otherwise we use the original `fd_read()`.
We also add a variant of `FS.createDevice()` called
`FS.createAsyncInputDevice()`.
Finally, if JSPI is available, we wrap the `main()` symbol with
`WebAssembly.promising()` so that we can stack switch from `fd_read()`.
If JSPI is not available, attempting to read from an AsyncInputDevice
will raise an `OSError`.
