Add a new _testinternalcapi module to test the internal C API.
Move _Py_GetConfigsAsDict() function to the internal C API:
_testembed now uses _testinternalcapi to access the function.
shutil.which() and distutils.spawn.find_executable() now use
os.confstr("CS_PATH") if available instead of os.defpath, if the PATH
environment variable is not set.
Don't use os.confstr("CS_PATH") nor os.defpath if the PATH
environment variable is set to an empty string to mimick Unix 'which'
command behavior.
Changes:
* find_executable() now starts by checking for the executable in the
current working directly case. Add an explicit
"if not path: return None".
* Add tests for PATH='' (empty string), PATH=':' and for PATHEXT.
IPv4Interface and IPv6Interface did not has netmask and hostmask
attributes when its argument is bytes or int.
This commit extracts method for constructors of Network and Interface,
and ensure Interface class always provides them.
The imap.IMAP4.logout() method no longer ignores silently arbitrary
exceptions.
Changes:
* The IMAP4.logout() method now expects a "BYE" untagged response,
rather than relying on _check_bye() which raises a self.abort()
exception.
* IMAP4.__exit__() now does nothing if the client already logged out.
* Add more debug info if test_logout() tests fail.
Add new keyword arguments "default_namespace" and "xml_declaration" to functions ET.tostring() and ET.tostringlist(), as known from ElementTree.write().
In Python having a trace function in effect while mock is imported causes isinstance to be wrong for MagicMocks. This is due to the usage of super() in some class methods, as this sets the __class__ attribute. To avoid this, as a workaround, alias the usage of super .
Omit serialno field from debug hooks on Python memory allocators to
reduce the memory footprint by 5%.
Enable tracemalloc to get the traceback where a memory block has been
allocated when a fatal memory error is logged to decide where to put
a breakpoint.
Compile Python with PYMEM_DEBUG_SERIALNO defined to get back the
field.
This is effectively an un-revert of #11617 and #12024 (reverted in #12159). Portions of those were merged in other PRs (with lower risk) and this represents the remainder. Note that I found 3 different bugs in the original PRs and have fixed them here.
In development mode (-X dev) and in debug build, the io.IOBase
destructor now logs close() exceptions. These exceptions are silent
by default in release mode.
Fix test_sys.test_getallocatedblocks() when tracemalloc is enabled.
If the name of Python memory allocators cannot get read, consider
that pymalloc is disabled.
Fix the following error:
./python -X tracemalloc -m test test_sys -v -m test_getallocatedblocks
ERROR: test_getallocatedblocks (test.test_sys.SysModuleTest)
------------------------------------------------------------
Traceback (most recent call last):
File "Lib/test/test_sys.py", line 770, in test_getallocatedblocks
alloc_name = _testcapi.pymem_getallocatorsname()
RuntimeError: cannot get allocators name
Modify CLEANBYTE, DEADDYTE and FORBIDDENBYTE constants: use 0xCD,
0xDD and 0xFD, rather than 0xCB, 0xBB and 0xFB, to use the same byte
patterns than Windows CRT debug malloc() and free().
Replace _PyMem_IsFreed() function with _PyMem_IsPtrFreed() inline
function. The function is now way more efficient, it became a simple
comparison on integers, rather than a short loop. It detects also
uninitialized bytes and "forbidden bytes" filled by debug hooks
on memory allocators.
Add unit tests on _PyObject_IsFreed().
The random module now prefers the lean internal _sha512 module over hashlib
for seed(version=2) to optimize import time.
Signed-off-by: Christian Heimes <christian@python.org>
Fixes some mistakes and misleadings in the quote function docstring:
- reserved chars are never actually used by quote code, unreserved chars are
- reserved chars were wrong and incomplete
- mentioned that use-case is not minimal quoting wrt. RFC, but cautious quoting
dash_R() function of libregrtest doesn't call support.gc_collect()
directly anymore: it's already called by dash_R_cleanup().
Call dash_R_cleanup() before starting the loop.
Fix reference leak hunting in regrtest: compute also deltas (of
reference count, allocated memory blocks, file descriptor count)
during warmup, to ensure that everything is initialized before
starting to hunt reference leaks.
Other changes:
* Replace gc.collect() with support.gc_collect()
* Move calls to read memory statistics from dash_R_cleanup() to
dash_R()
* Pass regrtest 'ns' to dash_R()
* dash_R() is now more quiet with --quiet option (don't display
progress).
* Precompute the full range for "for it in range(repcount):" to
ensure that the iteration doesn't allocate anything new.
* dash_R() now is responsible to call warm_caches().
While Windows exposes the system processor queue length, the raw value
used for load calculations on Unix systems, it does not provide an API
to access the averaged value. Hence to calculate the load we must track
and average it ourselves. We can't use multiprocessing or a thread to
read it in the background while the tests run since using those would
conflict with test_multiprocessing and test_xxsubprocess.
Thus, we use Window's asynchronous IO API to run the tracker in the
background with it sampling at the correct rate. When we wish to access
the load we check to see if there's new data on the stream, if there is,
we update our load values.
Move all documentation regarding the readinto method into either io.RawIOBase or io.BufferedIOBase.
Corresponding changes to documentation in the _pyio.py module.
It turns out doing socket.listen(0) does not equal to "choose a
reasonable default". It actually means "set backlog to 0".
As such set backlog=None as the default for socket.create_server.
Fixes the following BB failures:
https://github.com/python/cpython/pull/11784#issuecomment-481036369
Ref. BPO-1756, GH-11784.
* Properly handle SyntaxErrors in Python source files.
SyntaxErrors in the target module will rise normally, while SyntaxErrors in dependencies will be added to badmodules. This includes a new regression test.
* Fix name collision bug.
This fixes an issue where a "fromlist" import with the same name as a previously failed import would be incorrectly added to badmodules. This includes a new regression test.
* Replace mutable default values.
Bound empty lists have been replaced with the "if param is None" idiom.
* Replace deprecated imp usage.
Constants imported from imp have been moved to private module-level constants, and ModuleFinder.find_module has been refactored to use importlib. Other than an improvement on how frozen builtin imports are reported (as the frozen imports they are, rather than the stdlib modules they *may* have originated from), these changes maintain complete compatibility with past versions... including odd behavior for returning relative (below current directory, but not a C extension) vs. absolute (above current directory, or a C extension) paths.
Patch by Brandt Bucher.
Deprecated passing the following arguments as keyword arguments:
- "func" in functools.partialmethod(), weakref.finalize(),
profile.Profile.runcall(), cProfile.Profile.runcall(),
bdb.Bdb.runcall(), trace.Trace.runfunc() and
curses.wrapper().
- "function" in unittest.addModuleCleanup() and
unittest.TestCase.addCleanup().
- "fn" in the submit() method of concurrent.futures.ThreadPoolExecutor
and concurrent.futures.ProcessPoolExecutor.
- "callback" in contextlib.ExitStack.callback(),
contextlib.AsyncExitStack.callback() and
contextlib.AsyncExitStack.push_async_callback().
- "c" and "typeid" in the create() method of multiprocessing.managers.Server
and multiprocessing.managers.SharedMemoryServer.
- "obj" in weakref.finalize().
Also allowed to pass arbitrary keyword arguments (even "self" and "func")
if the above arguments are passed as positional argument.
Stop rejecting IPv4 octets with leading zeroes as ambiguously octal.
Plenty of other tools generate decimal IPv4 octets with leading zeroes,
so keeping this check hurts interoperability.
Patch by Joel Croteau.
Return None after calling unittest.mock.patch.object.stop() regardless of whether the object was started. This makes the method idempotent.
https://bugs.python.org/issue36366
bpo-36443, bpo-36202: Since Python 3.7.0, calling Py_DecodeLocale()
before Py_Initialize() produces mojibake if the LC_CTYPE locale is
coerced and/or if the UTF-8 Mode is enabled by the user
configuration. This change fix the issue by disabling LC_CTYPE
coercion and UTF-8 Mode by default. They must now be enabled
explicitly (opt-in) using the new _Py_PreInitialize() API with
_PyPreConfig.
When embedding Python, set coerce_c_locale and utf8_mode attributes
of _PyPreConfig to -1 to enable automatically these parameters
depending on the LC_CTYPE locale, environment variables and command
line arguments
Alternative: Setting Py_UTF8Mode to 1 always explicitly enables the
UTF-8 Mode.
Changes:
* _PyPreConfig_INIT now sets coerce_c_locale and utf8_mode to 0 by
default.
* _Py_InitializeFromArgs() and _Py_InitializeFromWideArgs() can now
be called with config=NULL.
* _PyCoreConfig_Write() now updates _PyRuntime.preconfig
* Remove _PyPreCmdline_Copy()
* _PyPreCmdline_Read() now accepts _PyPreConfig and _PyCoreConfig
optional configurations.
* Rename _PyPreConfig_ReadFromArgv() to _PyPreConfig_Read(). Simplify
the code.
* Calling _PyCoreConfig_Read() no longer adds the warning options
twice: don't add a warning option if it's already in the list.
* Rename _PyCoreConfig_ReadFromArgv() to _PyCoreConfig_Read().
* Rename config_from_cmdline() to _PyCoreConfig_ReadFromArgv().
* Add more assertions on _PyCoreConfig in _PyCoreConfig_Read().
* Move some functions.
* Make some config functions private.
* Add _Py_GetConfigsAsDict() function to get all configurations as a
dict.
* dump_config() of _testembed.c now dumps preconfig as a separated
key: call _Py_GetConfigsAsDict().
* Make _PyMainInterpreterConfig_AsDict() private.
Now that the parser generator is written in Python (Parser/pgen) we can make use of it to regenerate the Lib/keyword file that contains the language keywords instead of parsing the autogenerated grammar files. This also allows checking in the CI that the autogenerated files are up to date.
Before, an `AttributeError` was raised due to trying to access an attribute that exists on specs but having received `None` instead for a non-existent module.
https://bugs.python.org/issue36298
bpo-36256: Fix bug in parsermodule when parsing if statements
In the parser module, when validating nodes before starting the parsing with to create a ST in "parser_newstobject" there is a problem that appears when two arcs in the same DFA state has transitions with labels with the same type. For example, the DFA for if_stmt has a state with
two labels with the same type: "elif" and "else" (type NAME). The algorithm tries one by one the arcs until the label that starts the arc transition has a label with the same type of the current child label we are trying to accept. In this case, the arc for "elif" comes before the arc for "else"and passes this test (because the current child label is "else" and has the same type as "elif"). This lead to expecting a namedexpr_test (305) instead of a colon (11). The solution is to compare also the string representation (in case there is one) of the labels to see if the transition that we have is the correct one.
* bpo-35493: Use Process.sentinel instead of sleeping for polling worker status in multiprocessing.Pool
* Use self-pipe pattern to avoid polling for changes
* Refactor some variable names and add comments
* Restore timeout and poll
* Use reader object only on wait()
* Recompute worker sentinels every time
* Remove timeout and use change notifier
* Refactor some methods to be overloaded by the ThreadPool, document the cache class and fix typos
ProcessPoolExecutor workers will hold the return value of their last task in memory until the next task is received. Since the return value has already been propagated to the parent process's Future (or has been discarded by this point), the object can be safely released.
Fix CFLAGS in customize_compiler() of distutils.sysconfig: when the
CFLAGS environment variable is defined, don't override CFLAGS variable with
the OPT variable anymore.
Initial patch written by David Malcolm.
Co-Authored-By: David Malcolm <dmalcolm@redhat.com>
The value is a string for string and byte literals, None otherwise.
It is 'u' for u"..." literals, 'b' for b"..." literals, '' for "..." literals.
The 'r' (raw) prefix is ignored.
Does not apply to f-strings.
This appears sufficient to make mypy capable of using the stdlib ast module instead of typed_ast (assuming a mypy patch I'm working on).
WIP: I need to make the tests pass. @ilevkivskyi @serhiy-storchaka
https://bugs.python.org/issue36280
test_posix.PosixUidGidTests:
* Add tests for invalid uid/gid type (str)
* Add UID_OVERFLOW and GID_OVERFLOW constants to replace (1 << 32)
Initial patch written by David Malcolm.
Co-Authored-By: David Malcolm <dmalcolm@redhat.com>
* Refactor cookie path check as per RFC 6265
* Add tests for prefix match of path
* Add news entry
* Fix set_ok_path and refactor tests
* Use slice for last letter
Don't send cookies of domain A without Domain attribute to domain B when domain A is a suffix match of domain B while using a cookiejar with `http.cookiejar.DefaultCookiePolicy` policy. Patch by Karthikeyan Singaravelan.
The overflow check was relying on undefined behaviour as it was using the result of the multiplication to do the check, and once the overflow has already happened, any operation on the result is undefined behaviour.
Some extra checks that exercise code paths related to this are also added.
This adds a `feature_version` flag to `ast.parse()` (documented) and `compile()` (hidden) that allow tweaking the parser to support older versions of the grammar. In particular if `feature_version` is 5 or 6, the hacks for the `async` and `await` keyword from PEP 492 are reinstated. (For 7 or higher, these are unconditionally treated as keywords, but they are still special tokens rather than `NAME` tokens that the parser driver recognizes.)
https://bugs.python.org/issue35975
bpo-34247, bpo-36142: The PYTHONMALLOC environment variable has the
priority over PYTHONDEV env var and "-X dev" command line option.
For example, PYTHONMALLOC=malloc PYTHONDEVMODE=1 sets the memory
allocators to "malloc" (and not to "debug").
Add an unit test.
* _PyPreConfig_Write() now reallocates the pre-configuration with the
new memory allocator.
* It is no longer needed to force the "default raw memory allocator"
to clear pre-configuration and core configuration. Simplify the
code.
* _PyPreConfig_Write() now does nothing if called after
Py_Initialize(): no longer check if the allocator is the same.
* Remove _PyMem_GetDebugAllocatorsName(): dev mode sets again
allocator to "debug".
The development mode now uses the effective name of the debug memory
allocator ("pymalloc_debug" or "malloc_debug"). So the name doesn't
change after setting the memory allocator.
* Move 'allocator' and 'dev_mode' fields from _PyCoreConfig
to _PyPreConfig.
* Fix InitConfigTests of test_embed: dev_mode sets allocator to
"debug", add a new tests for env vars with dev mode enabled.
For C++ extensions, distutils tries to replace the C compiler with the
C++ compiler, but it assumes that C compiler is the first element after
any environment variables set. On AIX, linking goes through ld_so_aix,
so it is the first element and the compiler is the next element. Thus
the replacement is faulty:
ld_so_aix gcc ... -> g++ gcc ...
Also, it assumed that self.compiler_cxx had only 1 element or that
there were the same number of elements as the linker has and in the
same order. This might not be the case, so instead concatenate
everything together.
The previous code hardcoded `SEEK_SET`, etc. While it's very unlikely
that these values will change, it's best to use the definitions to avoid
there being mismatches in behavior with the code in the future.
Signed-off-by: Enji Cooper <yaneurabeya@gmail.com>
It is changed from 16KiB to 64KiB. The previous default value
is used since 1990.
coreutils chose 128 KiB as minimum buffer size for block device I/O.
But shutil.copyfileobj() can be used for non block devices.
So I choose more conservative value.
As my quick benchmark, performance difference between 64KiB and
128 KiB is up to ~5%. On the other hand, performance difference
between 32 KiB and 64 KiB can be more than 10% when file is fully
buffered.
This is why 64 KiB is rational value.
* Move fields from _PyMain to _PyCoreConfig:
* skip_first_line
* run_command
* run_module
* run_filename
* Replace _PyMain.stdin_is_interactive with a new
stdin_is_interactive(config) function
* Rename _PyMain to _PyArgv. Add "const _PyArgv *args" field
to _PyCmdline.
Use locale.getpreferredencoding() rather than locale.getlocale() to
get the locale encoding. With some locales, locale.getlocale()
returns the wrong encoding.
For example, on Fedora 29, locale.getlocale() returns ISO-8859-1
encoding for the "en_IN" locale, whereas
locale.getpreferredencoding() reports the correct encoding: UTF-8.
Responding to suggestions on the tracker and some off-line suggestions.
Davin suggested that english named accessors instead of greek letters would result in more intelligible user code. Steven suggested that the parameters still need to be *mu* and *theta* which are used elsewhere (and I noted those parameter names are used in linked-to resources).
Michael suggested proving-out the API by seeing whether it generalized to *Lognormal*. I did so and found that Lognormal distribution parameters *mu* and *sigma* do not represent the mean and standard deviation of the lognormal distribution (instead, they are for the underlying regular normal distribution).
Putting these ideas together, we have NormalDist parameterized by *mu* and *sigma* but offering English named properties for accessors. That gives lets us match other API that access mu and sigma, it matches the external resources on the topic, gives us clear english names in user code. The API extends nicely to LogNormal where the parameters and the summary statistic accessors are not the same.
https://bugs.python.org/issue36018
* Resolve string target to patch.dict during function call
* Add NEWS entry
* Remove unneeded call
* Restore original value for support.target and refactor assertions
* Add extra assertion to verify unpatched dict
* Added tests for shared_memory submodule.
* Added tests for ShareableList.
* Fix bug in allocationn size during creation of empty ShareableList illuminated by existing test run on Linux.
* Initial set of docs for shared_memory module.
* Added docs for ShareableList, added doctree entry for shared_memory submodule, name refactoring for greater clarity.
* Added examples to SharedMemoryManager docs, for ease of documentation switched away from exclusively registered functions to some explicit methods on SharedMemoryManager.
* Wording tweaks to docs.
* Fix test failures on Windows.
* Added tests around SharedMemoryManager.
* Documentation tweaks.
* Fix inappropriate test on Windows.
* Further documentation tweaks.
* Fix bare exception.
* Removed __copyright__.
* Fixed typo in doc, removed comment.
* Updated SharedMemoryManager preliminary tests to reflect change of not supporting all registered functions on SyncManager.
* Added Sphinx doctest run controls.
* CloseHandle should be in a finally block in case MapViewOfFile fails.
* Missed opportunity to use with statement.
* Switch to self.addCleanup to spare long try/finally blocks and save one indentation, change to use decorator to skip test instead.
* Simplify the posixshmem extension module.
Provide shm_open() and shm_unlink() functions. Move other
functionality into the shared_memory.py module.
* Added to doc around size parameter of SharedMemory.
* Changed PosixSharedMemory.size to use os.fstat.
* Change SharedMemory.buf to a read-only property as well as NamedSharedMemory.size.
* Marked as provisional per PEP411 in docstring.
* Changed SharedMemoryTracker to be private.
* Removed registered Proxy Objects from SharedMemoryManager.
* Removed shareable_wrap().
* Removed shareable_wrap() and dangling references to it.
* For consistency added __reduce__ to key classes.
* Fix for potential race condition on Windows for O_CREX.
* Remove unused imports.
* Update access to kernel32 on Windows per feedback from eryksun.
* Moved kernel32 calls to _winapi.
* Removed ShareableList.copy as redundant.
* Changes to _winapi use from eryksun feedback.
* Adopt simpler SharedMemory API, collapsing PosixSharedMemory and WindowsNamedSharedMemory into one.
* Fix missing docstring on class, add test for ignoring size when attaching.
* Moved SharedMemoryManager to managers module, tweak to fragile test.
* Tweak to exception in OpenFileMapping suggested by eryksun.
* Mark a few dangling bits as private as suggested by Giampaolo.
Trying to assign a value to __debug__ using the assignment operator is supposed to fail, but
a missing check for forbidden names when setting the context in the ast was preventing this behaviour.
Fix test_ssl for strict OpenSSL configuration like RHEL8 strict crypto policy.
Use older TLS version for minimum TLS version of the server SSL context if
needed, to test TLS version older than default minimum TLS version.
It checks that a SyntaxWarning is raised when compile specified
statement, that it is raised only once, that it is converted to
a SyntaxError when raised as exception, and that both warning and
exception objects have corresponding attributes.
Need to reset capturing groups between two SRE(match) callings in loops, this fixes wrong capturing groups in rare cases.
Also add a missing index in re.rst.
The error message emitted when returning invalid types from __fspath__ in interfaces that allow passing PathLike objects has been improved and now it does explain the origin of the error.
Not using `__class_getitem__()` fallback if there is a non-subcriptable metaclass was caused by a certain asymmetry between how `PySequenceMethods` and `PyMappingMethods` are used in `PyObject_GetItem`. This PR removes this asymmetry. No tests failed, so I assume it was not intentional.
It makes the existing smaller test more readable and robust at the same time.
The execution of a shell in interactive mode from CI and buildbot test automation wasn't working out. What would work locally in our terminals would only work within a fraction of automation systems. The integration test was a nice to have. painful. deleting. :)
* bpo-1054041: Exit properly by a signal after a ^C.
An uncaught KeyboardInterrupt exception means the user pressed ^C and
our code did not handle it. Programs that install SIGINT handlers are
supposed to reraise the SIGINT signal to the SIG_DFL handler in order
to exit in a manner that their calling process can detect that they
died due to a Ctrl-C. https://www.cons.org/cracauer/sigint.html
After this change on POSIX systems
while true; do python -c 'import time; time.sleep(23)'; done
can be stopped via a simple Ctrl-C instead of the shell infinitely
restarting a new python process.
What to do on Windows, or if anything needs to be done there has not
yet been determined. That belongs in its own PR.
TODO(gpshead): A unittest for this behavior is still needed.
* Do the unhandled ^C check after pymain_free.
* Return STATUS_CONTROL_C_EXIT on Windows.
* Fix ifdef around unistd.h include.
* 📜🤖 Added by blurb_it.
* Add STATUS_CTRL_C_EXIT to the os module on Windows
* Add unittests.
* Don't send CTRL_C_EVENT in the Windows test.
It was causing CI systems to bail out of the entire test suite.
See https://dev.azure.com/Python/cpython/_build/results?buildId=37980
for example.
* Correct posix test (fail on macOS?) check.
* STATUS_CONTROL_C_EXIT must be unsigned.
* Improve the error message.
* test typo :)
* Skip if the bash version is too old.
...and rename the windows test to reflect what it does.
* min bash version is 4.4, detect no bash.
* restore a blank line i didn't mean to delete.
* PyErr_Occurred() before the Py_DECREF(co);
* Don't add os.STATUS_CONTROL_C_EXIT as a constant.
* Update the Windows test comment.
* Refactor common logic into a run_eval_code_obj fn.
Previously, `debug print(` would cause the interpreter to exit on a SyntaxError whereas `print(` would properly display the error and return to the pdb prompt.
This patch fixes this by pre-compiling the code before passing it to `Pdb.run`.
https://bugs.python.org/issue35931
[bpo-35633](https://bugs.python.org/issue35633): Fix a test regression introduced with [bpo-35189](https://bugs.python.org/issue35189) (PEP 475: fnctl functions are not retried if interrupted (EINTR)).
Not only a blocking IO error needs to be ignored - permission errors also need to be ignored.
p.s. - iirc as a "test" only correction a NEWS item is not required. If this is not correct - just mention, and I'll add a NEWS blurb.
https://bugs.python.org/issue35633
Changes in this commit:
1. Use a _strong_ reference between the Pool and associated iterators
2. Rework PR #8450 to eliminate a cycle in the Pool.
There is no test in this commit because any test that automatically tests this behaviour needs to eliminate the pool before joining the pool to check that the pool object is garbaged collected/does not hang. But doing this will potentially leak threads and processes (see https://bugs.python.org/issue35413).
Protect dict iterations by wrapping them with _IterationGuard in the
following methods:
- WeakValueDictionary.copy()
- WeakValueDictionary.__deepcopy__()
- WeakKeyDictionary.copy()
- WeakKeyDictionary.__deepcopy__()
In http.server script, rely on getaddrinfo to bind to preferred address based on the bind parameter.
As a result, now IPv6 is used as the default (including IPv4 on dual-stack systems). Enhanced tests.