The function `operator.methodcaller` was not thread-safe since the additional
of the vectorcall method in gh-89013. In the free threading build the issue
is easy to trigger, for the normal build harder.
This makes the `methodcaller` safe by:
* Replacing the lazy initialization with initialization in the constructor.
* Using a stack allocated space for the vectorcall arguments and falling back
to `tp_call` for calls with more than 8 arguments.
Remove `PathBase.samefile()`, which is fairly specific to the local FS, and
relies on `stat()`, which we're aiming to remove from `PathBase`.
Also remove `PathBase.is_mount()`, `is_junction()`, `is_block_device()`,
`is_char_device()`, `is_fifo()` and `is_socket()`. These rely on POSIX
file type numbers that we're aiming to remove from the `PathBase` API.
Co-authored-by: Peter Bierma <zintensitydev@gmail.com>
Co-authored-by: Terry Jan Reedy <tjreedy@udel.edu>
Co-authored-by: Petr Viktorin <encukou@gmail.com>
Change the default value of `PurePathBase.parser` from `ParserBase()` to
`posixpath`. As a result, user subclasses of `PurePathBase` and `PathBase`
use POSIX path syntax by default, which is very often desirable.
Move `pathlib._abc.ParserBase` to `pathlib._types.Parser`, and convert it
to a runtime-checkable protocol.
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Adds a `use_system_log` config item to enable stdout/stderr redirection for
Apple platforms. This log streaming is then used by a new iOS test runner
script, allowing the display of test suite output at runtime. The iOS test
runner script can be used by any Python project, not just the CPython test
suite.
Virtual filesystems don't always make a distinction between deleting files
and empty directories, and sometimes support deleting non-empty directories
in a single operation. Here we remove `PathBase.unlink()` and `rmdir()`,
leaving `_delete()` as the sole deletion method, now made abstract. I hope
to drop the underscore prefix later on.
The canonical `file:` URL (as generated by `pathname2url()`) is now used as the `url` attribute of the returned `addinfourl` object. The `addinfourl.url` attribute reflects the resolved URL for both `file:` or `http[s]:` URLs now.
Remove our implementation of POSIX path resolution in `PathBase.resolve()`.
This functionality is rather fragile and isn't necessary in most cases. It
depends on `PathBase.stat()`, which we're looking to remove.
Also remove `PathBase.absolute()`. Many legitimate virtual filesystems lack
the notion of a 'current directory', so it's wrong to include in the basic
interface.
On Linux, threading.Thread now sets the thread name to the operating
system.
* configure now checks if pthread_getname_np()
and pthread_setname_np() functions are available.
* Add PYTHREAD_NAME_MAXLEN macro.
* Add _thread._NAME_MAXLEN constant for test_threading.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
In some cases, previously computed as (nan+nanj), we could recover
meaningful component values in the result, see e.g. the C11, Annex
G.5.1, routine _Cmultd():
>>> z = 1e300+1j
>>> z*(nan+infj) # was (nan+nanj)
(-inf+infj)
That also fix some complex powers for small integer exponents, computed
with optimized algorithm (by squaring):
>>> z**5 # was (nan+nanj)
Traceback (most recent call last):
File "<python-input-1>", line 1, in <module>
z**5
~^^~
OverflowError: complex exponentiation
Ensure `_SelectorSocketTransport.writelines` pauses the protocol if it reaches the high water mark as needed.
Co-authored-by: Kumar Aditya <kumaraditya@python.org>
Remove documentation for `pathlib.Path.scandir()`, and rename the method to
`_scandir()`. In the private pathlib ABCs, make `iterdir()` abstract and
call it from `_scandir()`.
It's not worthwhile to add this method at the moment - see discussion:
https://discuss.python.org/t/ergonomics-of-new-pathlib-path-scandir/71721
Co-authored-by: Steve Dower <steve.dower@microsoft.com>
gh-127572: Fix `test_structmembers` initialization.
The 'C' format code expects an `int` as a destination (not a `char`).
This led to test failures on big-endian platforms like s390x. Use the
'c' format code, which expects a `char` as the destination (but requires
a Python byte objects instead of a str).
The CALL family of instructions were mostly thread-safe already and only required a small number of changes, which are documented below.
A few changes were needed to make CALL_ALLOC_AND_ENTER_INIT thread-safe:
Added _PyType_LookupRefAndVersion, which returns the type version corresponding to the returned ref.
Added _PyType_CacheInitForSpecialization, which takes an init method and the corresponding type version and only populates the specialization cache if the current type version matches the supplied version. This prevents potentially caching a stale value in free-threaded builds if we race with an update to __init__.
Only cache __init__ functions that are deferred in free-threaded builds. This ensures that the reference to __init__ that is stored in the specialization cache is valid if the type version guard in _CHECK_AND_ALLOCATE_OBJECT passes.
Fix a bug in _CREATE_INIT_FRAME where the frame is pushed to the stack on failure.
A few other miscellaneous changes were also needed:
Use {LOCK,UNLOCK}_OBJECT in LIST_APPEND. This ensures that the list's per-object lock is held while we are appending to it.
Add missing co_tlbc for _Py_InitCleanup.
Stop/start the world around setting the eval frame hook. This allows us to read interp->eval_frame non-atomically and preserves the behavior of _CHECK_PEP_523 documented below.
* Replace uses of `PyCell_GET` and `PyCell_SET`. These macros are not
safe to use in the free-threaded build. Use `PyCell_GetRef()` and
`PyCell_SetTakeRef()` instead.
* Since `PyCell_GetRef()` returns a strong rather than borrowed ref, some
code restructuring was required, e.g. `frame_get_var()` returns a strong
ref now.
* Add critical sections to `PyCell_GET` and `PyCell_SET`.
* Move critical_section.h earlier in the Python.h file.
* Add `PyCell_GET` to the free-threading howto table of APIs that return
borrowed refs.
* Add additional unit tests for free-threading.
No additional thread safety changes are required. Note that sending to
a generator that is shared between threads is currently not safe in the
free-threaded build.
Fix race in test_start_new_thread_failed
When we succeed in starting a new thread, for example if setrlimit
was ineffective, we must wait for the newly spawned thread to exit.
Otherwise, we run the risk that the newly spawned thread will race
with runtime finalization and access memory that has already been
clobbered/freed.
`_thread.start_new_thread()` only spawns daemon threads, which the runtime
does not wait for at shutdown, and does not return a handle. Use
`_thread.start_joinable_thread()` and join the resulting handle when
the thread is started successfully.
Use existing helpers to atomically modify the bytecode. Add unit tests
to ensure specializing is happening as expected. Add test_specialize.py
that can be used with ThreadSanitizer to detect data races.
Fix thread safety issue with cell_set_contents().
Previously, `traceback.print_list` didn't have a documentation entry and was not exposed in `traceback.__all__`. Now it has a documentation entry and is exposed in `__all__`.
* Mark almost all reachable objects before doing collection phase
* Add stats for objects marked
* Visit new frames before each increment
* Update docs
* Clearer calculation of work to do.
These classmethods presume that the user has retained the original
`__init__()` signature, which may not be the case. Also, many virtual
filesystems don't provide current or home directories.
Since MultiByteToWideChar()/WideCharToMultiByte() is not reversible if
the data contains invalid UTF-8 sequences, use binary search to
calculate the number of written bytes from the number of written
characters.
Also fix writing incomplete UTF-8 sequences.
Also fix handling of memory allocation failures.
"Generally, mixed-mode arithmetic combining real and complex variables should
be performed directly, not by first coercing the real to complex, lest the sign
of zero be rendered uninformative; the same goes for combinations of pure
imaginary quantities with complex variables." (c) Kahan, W: Branch cuts for
complex elementary functions.
This patch implements mixed-mode arithmetic rules, combining real and
complex variables as specified by C standards since C99 (in particular,
there is no special version for the true division with real lhs
operand). Most C compilers implementing C99+ Annex G have only these
special rules (without support for imaginary type, which is going to be
deprecated in C2y).
* Revert "[3.13] gh-124873: Tolerate 100 ms in TimerfdTests on Android (GH-127101) (#127105)"
This reverts commit c09366b1fe.
* Skip timerfd tests on Android.
Co-authored-by: Victor Stinner <vstinner@python.org>
When handed an absolute Windows path such as `C:\foo` or `//server/share`,
the `urllib.request.pathname2url()` function returns a URL with an
authority section, such as `///C:/foo` or `//server/share` (or before
GH-126205, `////server/share`). Only the `file:` prefix is omitted.
But when handed an absolute POSIX path such as `/etc/hosts`, or a Windows
path of the same form (rooted but lacking a drive), the function returns a
URL without an authority section, such as `/etc/hosts`.
This patch corrects the discrepancy by adding a `//` prefix before
drive-less, rooted paths when generating URLs.
Move tests for Path.walk() into a new PathWalkTest class, and apply a similar change in tests for the ABCs. This allows us to properly tear down the walk test hierarchy in tearDown(), rather than leaving it to os_helper.rmtree().
Stop converting Windows drive letters to uppercase in
`urllib.request.pathname2url()` and `url2pathname()`. This behaviour is
unnecessary and inconsistent with pathlib's file URI implementation.
If Python fails to start newly created thread
due to failure of underlying PyThread_start_new_thread() call,
its state should be removed from interpreter' thread states list
to avoid its double cleanup.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
Add free-threaded specialization for `UNPACK_SEQUENCE` opcode.
`UNPACK_SEQUENCE_TUPLE/UNPACK_SEQUENCE_TWO_TUPLE` are already thread safe since tuples are immutable.
`UNPACK_SEQUENCE_LIST` is not thread safe because of nature of lists (there is nothing preventing another thread from adding items to or removing them the list while the instruction is executing). To achieve thread safety we add a critical section to the implementation of `UNPACK_SEQUENCE_LIST`, especially around the parts where we check the size of the list and push items onto the stack.
---------
Co-authored-by: Matt Page <mpage@meta.com>
Co-authored-by: mpage <mpage@cs.stanford.edu>
Threads are gone after fork, so clear the queues too. Otherwise the
child process (here created via multiprocessing.Process) crashes on
interpreter exit.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
* Name without a PATHEXT extension is only searched if the mode does not
include X_OK.
* Support multi-component PATHEXT extensions (e.g. ".foo.bar").
* Support files without extensions in PATHEXT contains dot-only extension
(".", "..", etc).
* Support PATHEXT extensions that end with a dot (e.g. ".foo.").
The usage parameter of argparse.ArgumentParser no longer
affects the default value of the prog parameter in subparsers.
Previously the full custom usage of the main parser was used as
the prog prefix in subparsers.
`mmap`, `munmap`, and `mprotect` are used by CPython for memory
management, which may occur in the middle of the FileIO tests. The
system calls can also be used with files, so `strace` includes them
in its `%file` and `%desc` filters.
Filter out the `mmap` system calls related to memory allocation for the
file tests. Currently FileIO doesn't do `mmap` at all, so didn't add
code to track from `mmap` through `munmap` since it wouldn't be used.
For now if an `mmap` on a fd happens, the call will be included (which
may cause test to fail), and at that time support for tracking the
address throug `munmap` could be added.
Decode a file URI like `file://///server/share` as a UNC path like
`\\server\share`. This form of file URI is created by software the simply
prepends `file:///` to any absolute Windows path.
Discard any 'localhost' authority from the beginning of a `file:` URI. As a
result, file URIs like `//localhost/etc/hosts` are correctly decoded as
`/etc/hosts`.
Adjust `pathname2url()` to encode embedded colon characters in Windows
paths, rather than bailing out with an `OSError`.
Co-authored-by: Steve Dower <steve.dower@microsoft.com>
Enable specialization of LOAD_GLOBAL in free-threaded builds.
Thread-safety of specialization in free-threaded builds is provided by the following:
A critical section is held on both the globals and builtins objects during specialization. This ensures we get an atomic view of both builtins and globals during specialization.
Generation of new keys versions is made atomic in free-threaded builds.
Existing helpers are used to atomically modify the opcode.
Thread-safety of specialized instructions in free-threaded builds is provided by the following:
Relaxed atomics are used when loading and storing dict keys versions. This avoids potential data races as the dict keys versions are read without holding the dictionary's per-object lock in version guards.
Dicts keys objects are passed from keys version guards to the downstream uops. This ensures that we are loading from the correct offset in the keys object. Once a unicode key has been stored in a keys object for a combined dictionary in free-threaded builds, the offset that it is stored in will never be reused for a different key. Once the version guard passes, we know that we are reading from the correct offset.
The dictionary read fast-path is used to read values from the dictionary once we know the correct offset.
* Fix support of STRING and GLOBAL opcodes with non-ASCII arguments.
* dis() now outputs non-ASCII bytes in STRING, BINSTRING and
SHORT_BINSTRING arguments as escaped (\xXX).
Distribution tooling (ex. sandbox on Gentoo and fakeroot on Debian) uses
LD_PRELOAD to intercept system calls and potentially modify them when
building. These tools can change the set of system calls, so disable
system call testing under these cases.
Co-authored-by: Michał Górny <mgorny@gentoo.org>
This adds authentication to the forkserver control socket. In the past only filesystem permissions protected this socket from code injection into the forkserver process by limiting access to the same UID, which didn't exist when Linux abstract namespace sockets were used (see issue) meaning that any process in the same system network namespace could inject code. We've since stopped using abstract namespace sockets by default, but protecting our control sockets regardless of type is a good idea.
This reuses the HMAC based shared key auth already used by `multiprocessing.connection` sockets for other purposes.
Doing this is useful so that filesystem permissions are not relied upon and trust isn't implied by default between all processes running as the same UID with access to the unix socket.
### pyperformance benchmarks
No significant changes. Including `concurrent_imap` which exercises `multiprocessing.Pool.imap` in that suite.
### Microbenchmarks
This does _slightly_ slow down forkserver use. How much so appears to depend on the platform. Modern platforms and simple platforms are less impacted. This PR adds additional IPC round trips to the control socket to tell forkserver to spawn a new process. Systems with potentially high latency IPC are naturally impacted more.
Typically a 1-4% slowdown on a very targeted process creation microbenchmark, with a worst case overloaded system slowdown of 20%. No evidence that these slowdowns appear in practical sense. See the PR for details.
Adjust `urllib.request.url2pathname()` and `pathname2url()` to use the
filesystem encoding when quoting and unquoting file URIs, rather than
forcing use of UTF-8.
No changes are needed in the `nturl2path` module because Windows always
uses UTF-8, per PEP 529.
This approach eliminates the originally reported race. It also gets rid of the deadlock reported in gh-96071, so we can remove the workaround added then.
If the cpXXX encoding is not directly implemented in Python, fall back
to use the Windows-specific API codecs.code_page_encode() and
codecs.code_page_decode().
* Mark almost all reachable objects before doing collection phase
* Add stats for objects marked
* Visit new frames before each increment
* Remove lazy dict tracking
* Update docs
* Clearer calculation of work to do.
Modify the extended attribute tests to write fewer and smaller extended
attributes, in order to fit within filesystems with total xattr limit
of 1 KiB (e.g. ext4 with 1 KiB blocks). Previously, the test would
write over 2 KiB, making it fail with ENOSPC on such systems.
If SRE(match) function terminates abruptly, either because of a signal
or because memory allocation fails, allocated SRE_REPEAT blocks might
be never released.
Co-authored-by: <wjssz@users.noreply.github.com>
The header-folder of the new email API has a long standing known buglet where
if the first token is longer than max_line_length, it puts that token on the next
line. It turns out there is also a *parsing* bug when parsing such a header:
the space prefixing that first, non-empty line gets preserved and tacked on to
the start of the header value, which is not the expected behavior per the RFCs.
The bug arises from the fact that the parser assumed that there would be at
least one token on the line with the header, which is going to be true for
probably every email producer other than the python email library with its
folding buglet. Clearly, though, this is a case that needs to be handled
correctly. The fix is simple: strip the blanks off the start of the whole
value, not just the first physical line of the value.
Co-authored-by: blurb-it[bot] <43283697+blurb-it[bot]@users.noreply.github.com>
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
* Document that slices can be marshalled
* Deduplicate and organize the list of supported types
in docs
* Organize the type code list in marshal.c, to make
it more obvious that this is a versioned format
* Back-fill some historical info
Co-authored-by: Michael Droettboom <mdboom@gmail.com>
For dlsym(), a return value of NULL does not necessarily indicate
an error [1].
Therefore, to avoid using stale (or NULL) dlerror() values, we must:
1. clear the previous error state by calling dlerror()
2. call dlsym()
3. call dlerror()
If the return value of dlerror() is not NULL, an error occured.
In ctypes we choose to treat a NULL return value from dlsym()
as a "not found" error. This is the same as the fallback
message we use on Windows, Cygwin or when getting/formatting
the error reason fails.
[1]: https://man7.org/linux/man-pages/man3/dlsym.3.html
Signed-off-by: Georgios Alexopoulos <grgalex42@gmail.com>
Signed-off-by: Georgios Alexopoulos <grgalex@ba.uoa.gr>
Co-authored-by: Peter Bierma <zintensitydev@gmail.com>
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Co-authored-by: Petr Viktorin <encukou@gmail.com>
Fixes a bug where pygettext would attempt
to extract a message from a code like this:
def _(x): pass
This is because pygettext only looks at one
token at a time and '_(x)' looks like a
function call.
However, since 'x' is not a string literal,
it would erroneously issue a warning.
Discard two leading slashes from the beginning of a `file:` URI if they
introduce an empty authority section. As a result, file URIs like
`///etc/hosts` are correctly parsed as `/etc/hosts`.
In strict mode, raise `NotADirectoryError` if we encounter a non-directory
while we still have path parts left to process.
We use a `part_count` variable rather than `len(rest)` because the `rest`
stack also contains markers for unresolved symlinks.
Previously, this would cause an `AttributeError` if the patch stopped more than once after this, and would also disrupt the original patched object.
---------
Co-authored-by: Peter Bierma <zintensitydev@gmail.com>
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Accepting objects with false values (like 0 and []) except empty strings
and byte-like objects and None in urllib.parse functions parse_qsl() and
parse_qs() is now deprecated.
Co-authored-by: blurb-it[bot] <43283697+blurb-it[bot]@users.noreply.github.com>
Co-authored-by: Hugo van Kemenade <1324225+hugovk@users.noreply.github.com>
* Removes erroneous explanation of the `global` statement restrictions; a name declared as global can be subsequently bound using any kind of name binding operation.
* Updates `test_global.py` to also test various name-binding scenarios for global
variables to ensure correct behavior
The first version had it running two forkserver and one spawn tests underneath each of the _fork, _forkserver, and _spawn test suites that build off the generic one.
This adds to the existing complexity of the multiprocessing test suite by offering BaseTestCase classes another attribute to control which suites they are invoked under. Practicality vs purity here. :/
Net result: we don't over-run the new test and their internal logic is simplified.
Skip `testMakefileCloseSocketDestroy` test if `sys.getrefcount` isn't available. This is necessary for PyPy and other Python implementations that do not have `sys.getrefcount`.
gh-117378: Fix multiprocessing forkserver preload sys.path inheritance.
`sys.path` was not properly being sent from the parent process when launching
the multiprocessing forkserver process to preload imports. This bug has been
there since the forkserver start method was introduced in Python 3.4. It was
always _supposed_ to inherit `sys.path` the same way the spawn method does.
Observable behavior change: A `''` value in `sys.path` will now be replaced in
the forkserver's `sys.path` with an absolute pathname
`os.path.abspath(os.getcwd())` saved at the time that `multiprocessing` was
imported in the parent process as it already was when using the spawn start
method. **This will only be observable during forkserver preload imports**.
The code invoked before calling things in another process already correctly sets `sys.path`.
Which is likely why this went unnoticed for so long as a mere performance issue in
some configurations.
A workaround for the bug on impacted Pythons is to set PYTHONPATH in the
environment before multiprocessing's forkserver process was started. Not perfect
as that is then inherited by other children, etc, but likely good enough for many
people's purposes.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
In `PathBase.resolve()`, raise `UnsupportedOperation` if a non-POSIX path
parser is used (our implementation uses `posixpath._realpath()`, which
produces incorrect results for non-POSIX path flavours.) Also tweak code to
call `self.absolute()` upfront rather than supplying an emulated `getcwd()`
function.
Adjust `PathBase.absolute()` to work somewhat like `resolve()`. If a POSIX
path parser is used, we treat the root directory as the current directory.
This is the simplest useful behaviour for concrete path types without a
current directory cursor.
In the past I've equivocated about whether to require at least one argument
in the `PurePathBase` (and `PathBase`) initializer, and what the default
should be if we make it optional. I now have a local use case that has
persuaded me to make it optional and default to the empty string (a
`zipp.Path`-like class that treats relative and absolute paths similarly.)
Happily this brings the base class more in line with `PurePath` and `Path`.
Adjust `urllib.request.pathname2url()` and `url2pathname()` so that they
don't remove slashes from Windows DOS drive paths and URLs. There was no
basis for this behaviour, and it conflicts with how UNC and POSIX paths are
handled.
Move creation of a tuple for var-positional parameter out of
_PyArg_UnpackKeywordsWithVararg().
Merge _PyArg_UnpackKeywordsWithVararg() with _PyArg_UnpackKeywords().
Add a new parameter in _PyArg_UnpackKeywords().
The "parameters" and "converters" attributes of ParseArgsCodeGen no
longer contain the var-positional parameter. It is now available as the
"varpos" attribute. Optimize code generation for var-positional
parameter and reuse the same generating code for functions with and without
keyword parameters.
Add special converters for var-positional parameter. "tuple" represents it as
a Python tuple and "array" represents it as a continuous array of PyObject*.
"object" is a temporary alias of "tuple".
* gh-126298: Don't deduplicated slice constants based on equality
* NULL check for PySlice_New
* Fix refcounting
* Fix refcounting some more
* Fix refcounting
* Make tests more complete
* Fix tests
If `read()` in the ConnectionHandler thread raises `OSError` (except `ConnectionError`),
the ConnectionHandler shuts down the entire ThreadedEchoServer,
preventing further connections.
It also does that for `EPROTOTYPE` in `wrap_conn`.
As far as I can see, this is done to avoid the server thread getting stuck,
forgotten, in its accept loop. However, since 2011 (5b95eb90a7)
the server is used as a context manager, and its `__exit__` does `stop()` and `join()`.
(I'm not sure if we *always* used `with` since that commit, but currently we do.)
Make sure that the context manager *is* used, and remove the `server.stop()`
calls from ConnectionHandler.
The skipping machinery called `getattr(err, "reason", "")` on an arbitrary
exception. As intermittent Buildbot failures show, sometimes it's set
to None.
Convert it to string for this specific check.
- The specialization logic determines the appropriate specialization using only the operand's type, which is safe to read non-atomically (changing it requires stopping the world). We are guaranteed that the type will not change in between when it is checked and when we specialize the bytecode because the types involved are immutable (you cannot assign to `__class__` for exact instances of `dict`, `set`, or `frozenset`). The bytecode is mutated atomically using helpers.
- The specialized instructions rely on the operand type not changing in between the `DEOPT_IF` checks and the calls to the appropriate type-specific helpers (e.g. `_PySet_Contains`). This is a correctness requirement in the default builds and there are no changes to the opcodes in the free-threaded builds that would invalidate this.
Defer joining of path segments in the private `PurePathBase` ABC. The new
behaviour matches how the public `PurePath` class handles path segments.
This removes a hard-to-grok difference between the ABCs and the main
classes. It also slightly reduces the size of `PurePath` objects by
eliminating a `_raw_path` slot.
Replace the os.environ.refresh() method with a new
os.reload_environ() function.
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Co-authored-by: Adam Turner <9087854+AA-Turner@users.noreply.github.com>
Each thread specializes a thread-local copy of the bytecode, created on the first RESUME, in free-threaded builds. All copies of the bytecode for a code object are stored in the co_tlbc array on the code object. Threads reserve a globally unique index identifying its copy of the bytecode in all co_tlbc arrays at thread creation and release the index at thread destruction. The first entry in every co_tlbc array always points to the "main" copy of the bytecode that is stored at the end of the code object. This ensures that no bytecode is copied for programs that do not use threads.
Thread-local bytecode can be disabled at runtime by providing either -X tlbc=0 or PYTHON_TLBC=0. Disabling thread-local bytecode also disables specialization.
Concurrent modifications to the bytecode made by the specializing interpreter and instrumentation use atomics, with specialization taking care not to overwrite an instruction that was instrumented concurrently.
Mark Shannon