This roughly follows what was done for dictobject to make a lock-free
lookup operation. With this change, the set contains operation scales much
better when used from multiple-threads. The frozenset contains performance
seems unchanged (as already lock-free).
Summary of changes:
* refactor set_lookkey() into set_do_lookup() which now takes a function
pointer that does the entry comparison. This is similar to dictobject and
do_lookup(). In an optimized build, the comparison function is inlined and
there should be no performance cost to this.
* change set_do_lookup() to return a status separately from the entry value
* add set_compare_frozenset() and use if the object is a frozenset. For the
free-threaded build, this avoids some overhead (locking, atomic operations,
incref/decref on key)
* use FT_ATOMIC_* macros as needed for atomic loads and stores
* use a deferred free on the set table array, if shared (only on free-threaded
build, normal build always does an immediate free)
* for free-threaded build, use explicit for loop to zero the table, rather than memcpy()
* when mutating the set, assign so->table to NULL while the change is a
happening. Assign the real table array after the change is done.
Makes the zlib module thread-safe free-threading build. Even though operations
are protected by locks, attributes exposed via PyMemberDef (eof, needs_input,
unused_data, unconsumed_tail) should still be stored atomically within locked
sections, since they can be read without acquiring the lock.
If there are many untracked tuples, the GC will run too often, resulting
in poor performance. The fix is to include untracked tuples in the
"long lived" object count. The number of frozen objects is also now
included since the free-threaded GC must scan those too.
* Pick target depending on preconditions
* Clarify the news fragment
* Add test capturing missed expectation.
* Add more idiomatic safe realpath helper
* Restore logic where existance and directoriness are checked on realpath.
* Link GH issue to test.
* Extract a function to check the target. Remove the _safe_realpath, now no longer needed.
* Extract method for replacing sys_path, and isolate realpath usage there.
* Revert "Extract method for replacing sys_path, and isolate realpath usage there."
This reverts commit 855aac3d28.
* Restore _safe_realpath.
---------
Co-authored-by: Jason R. Coombs <jaraco@jaraco.com>
For optional arguments with required=True, the ArgumentDefaultsHelpFormatter
would always add a " (default: None)" to the end of the help text.
Since that's a bit misleading, it is removed with this commit.
* bpo-40350: fix namespace package support in modulefinder
Signed-off-by: Filipe Laíns <lains@riseup.net>
* consider that namespace package specs might have the loader set
Signed-off-by: Filipe Laíns <lains@riseup.net>
* Revert "consider that namespace package specs might have the loader set"
This reverts commit 23fb4e0de3.
Signed-off-by: Filipe Laíns <lains@riseup.net>
* Fix load_module and load_package to handle namespace packages
Signed-off-by: Filipe Laíns <lains@riseup.net>
* Drop _NAMESPACE constant
Signed-off-by: Filipe Laíns <lains@riseup.net>
* Drop importlib changes
Signed-off-by: Filipe Laíns <lains@riseup.net>
* Update NamespacePath check
Signed-off-by: Filipe Laíns <lains@riseup.net>
* Mixed some
Signed-off-by: Filipe Laíns <lains@riseup.net>
---------
Signed-off-by: Filipe Laíns <lains@riseup.net>
The tests were flaky on slow machines because subprocesses could finish
before enough samples were collected. This adds synchronization similar
to test_external_inspection: test scripts now signal when they start
working, and the profiler waits for this signal before sampling.
Test scripts now run in infinite loops until killed rather than for
fixed iterations, ensuring the profiler always has active work to
sample regardless of machine speed.
This PR implements frame caching in the RemoteUnwinder class to significantly reduce memory reads when profiling remote processes with deep call stacks.
When cache_frames=True, the unwinder stores the frame chain from each sample and reuses unchanged portions in subsequent samples. Since most profiling samples capture similar call stacks (especially the parent frames), this optimization avoids repeatedly reading the same frame data from the target process.
The implementation adds a last_profiled_frame field to the thread state that tracks where the previous sample stopped. On the next sample, if the current frame chain reaches this marker, the cached frames from that point onward are reused instead of being re-read from remote memory.
The sampling profiler now enables frame caching by default.
Extend defect handling via policy to a couple of missed defects.
---------
Co-authored-by: Martin Panter <vadmium@users.noreply.github.com>
Co-authored-by: Ivo Bellin Salarin <ivo@nilleb.com>
The keyword typo suggestion mechanism in traceback would incorrectly
suggest replacements when the extracted source code was merely incomplete
rather than containing an actual typo. For example, when a missing comma
caused a syntax error, the system would suggest replacing 'print' with
'not' because the incomplete code snippet happened to pass validation.
The fix adds a validation step that first checks whether the original
extracted code raises a SyntaxError. If the code compiles successfully
or is simply incomplete (compile_command returns None), the function
returns early since there is no way to verify that a keyword replacement
would actually fix the problem.
RDM: This fixes a subtle folding error that showed up when a token exactly filled a line and was followed by whitespace and a token with no folding whitespace that was longer than a line. In this particular circumstance the whitespace after the first token got pushed on to the next line, and then stolen to go in front of the next unfoldable token...leaving a completely empty line in the line buffer. That line got turned in to a newline, which is RFC illegal, and the newish security check caught it. The fix is to just delete that empty line from the buffer.
Co-authored-by: blurb-it[bot] <43283697+blurb-it[bot]@users.noreply.github.com>
Enhanced the _find_lineno method in doctest to correctly identify and
report line numbers for doctests defined in __test__ dictionaries when
formatted as triple-quoted strings.
Finds a non-blank line in the test string and matches it in the source
file, verifying subsequent lines also match to handle duplicate lines.
Previously, doctest would report "line None" for __test__ dictionary
strings, making it difficult to debug failing tests.
Co-authored-by: Jurjen N.E. Bos <jneb@users.sourceforge.net>
Co-authored-by: R. David Murray <rdmurray@bitdance.com>
