[3.9] bpo-41060: Avoid SEGFAULT when calling GET_INVALID_TARGET in the grammar (GH-21020) (GH-21024)
`GET_INVALID_TARGET` might unexpectedly return `NULL`, which if not
caught will cause a SEGFAULT. Therefore, this commit introduces a new
inline function `RAISE_SYNTAX_ERROR_INVALID_TARGET` that always
checks for `GET_INVALID_TARGET` returning NULL and can be used in
the grammar, replacing the long C ternary operation used till now.
Related to https://github.com/pypa/twine/issues/638 and https://github.com/pypa/packaging.python.org/issues/730, I wrote a spec based on the one that was removed in https://github.com/python/cpython/pull/13087. However, a Google search for "pypirc" turned up at least one [blog post](https://truveris.github.io/articles/configuring-pypirc/) that links to https://docs.python.org/3/distutils/packageindex.htmlGH-the-pypirc-file, which now just links to this document. So, I thought a link to the spec would be handy.
bpo-38377: Fix skip_if_broken_multiprocessing_synchronize() on macOS (GH-20984)
skip_if_broken_multiprocessing_synchronize() only attempts for create
a semaphore on Linux to fix multiprocessing
test_resource_tracker_reused() on macOS.
(cherry picked from commit 3358da4054b9b0b045eb47dc74dee3d58bfbb1d5)
Co-authored-by: Victor Stinner <vstinner@python.org>
[3.9] bpo-40334: Produce better error messages on invalid targets (GH-20106) (GH-20973)
* bpo-40334: Produce better error messages on invalid targets (GH-20106)
The following error messages get produced:
- `cannot delete ...` for invalid `del` targets
- `... is an illegal 'for' target` for invalid targets in for
statements
- `... is an illegal 'with' target` for invalid targets in
with statements
Additionally, a few `cut`s were added in various places before the
invocation of the `invalid_*` rule, in order to speed things
up.
On Linux, skip tests using multiprocessing if the current user cannot
create a file in /dev/shm/ directory. Add the
skip_if_broken_multiprocessing_synchronize() function to the
test.support module.
Inada Naoki [Thu, 18 Jun 2020 08:30:53 +0000 (17:30 +0900)]
bpo-36346: Add Py_DEPRECATED to deprecated unicode APIs (GH-20878)
Co-authored-by: Kyle Stanley <aeros167@gmail.com> Co-authored-by: Victor Stinner <vstinner@python.org>
(cherry picked from commit 2c4928d37edc5e4aeec3c0b79fa3460b1ec9b60d)
stratakis [Wed, 17 Jun 2020 19:43:27 +0000 (21:43 +0200)]
[3.9] bpo-40637: Do not emit warnings for disabled builtin hashes (GH… (GH-20951)
…-20937)
test_hashlib emits some warnings when it cannot find some hashes
as it assumes they failed to compile. Since we can disable hashes
through configure, we emit the warnings only in the case that we
did not intentionaly disable them.
Automerge-Triggered-By: @tiran
(cherry picked from commit 236a0f5)
Victor Stinner [Wed, 17 Jun 2020 16:40:14 +0000 (18:40 +0200)]
bpo-41003: Fix test_copyreg when numpy is installed (GH-20935) (GH-20945)
Fix test_copyreg when numpy is installed: test.pickletester now
saves/restores warnings.filters when importing numpy, to ignore
filters installed by numpy.
Add the save_restore_warnings_filters() function to the
test.support.warnings_helper module.
bpo-40993: Don't run Travis CI coverage on PRs (GH-20916)
C and Python coverage jobs of Travis CI are no longer run on pull
requests, only on branches like master.
(cherry picked from commit fc710ee266e9461fdba9933ec6004318db588820)
Co-authored-by: Victor Stinner <vstinner@python.org>
bpo-40985: Show correct SyntaxError text when last line has a LINECONT (GH-20888)
When a file ends with a line that contains a line continuation character
the text of the emitted SyntaxError is empty, contrary to the old
parser, where the error text contained the text of the last line.
(cherry picked from commit 113e2b0a07c72c0d5e3489076afb14f6b3ad1049)
bpo-36020: Remove snprintf macro in pyerrors.h (GH-20889)
On Windows, GH-include "pyerrors.h" no longer defines "snprintf" and
"vsnprintf" macros.
PyOS_snprintf() and PyOS_vsnprintf() should be used to get portable
behavior.
Replace snprintf() calls with PyOS_snprintf() and replace vsnprintf()
calls with PyOS_vsnprintf().
(cherry picked from commit e822e37946f27c09953bb5733acf3b07c2db690f)
Co-authored-by: Victor Stinner <vstinner@python.org>
ensurepip optionally installs or upgrades 'pip' and 'setuptools' using
the version of those modules bundled with Python. The internal PIP
installation routine by default temporarily uses its cache, if it
exists. This is undesirable as Python builds and installations may be
independent of the user running the build, whilst PIP cache location
is dependent on the user's environment and outside of the build
environment.
At the same time, there's no value in using the cache while installing
bundled modules.
bpo-37556 Extend help to include latest overrides (GH-14701)
Modify the help in cpython/PC/launcher.c to show users that "latest" can be overridden by shebang, PY_PYTHON[n] or py.ini files. Also show that script [args] is optional by enclosing in square brackets.
bpo-34226: fix cgi.parse_multipart without content_length (GH-8530)
In Python 3.7 the behavior of parse_multipart changed requiring CONTENT-LENGTH
header, this fix remove this header as required and fix FieldStorage
read_lines_to_outerboundary, by not using limit when it's negative,
since by default it's -1 if not content-length and keeps substracting what
was read from the file object.
bpo-40955: Fix memory leak in subprocess module (GH-20825)
```
Direct leak of 8 byte(s) in 1 object(s) allocated from:
GH-0 0x7f008bf19667 in __interceptor_malloc (/lib64/libasan.so.6+0xb0667)
GH-1 0x7f007a0bee4a in subprocess_fork_exec /home/heimes/dev/python/cpython/Modules/_posixsubprocess.c:774
GH-2 0xe0305b in cfunction_call Objects/methodobject.c:546
```
bpo-40847: Consider a line with only a LINECONT a blank line (GH-20769)
A line with only a line continuation character should be considered
a blank line at tokenizer level so that only a single NEWLINE token
gets emitted. The old parser was working around the issue, but the
new parser threw a `SyntaxError` for valid input. For example,
an empty line following a line continuation character was interpreted
as a `SyntaxError`.
bpo-40895: Update weakref documentation to remove old warnings (GH-20687)
The doccumentation at https://docs.python.org/3.10/library/weakref.html cautions that the `WeakKeyDictionary` and `WeakValueDictionary` are susceptible to the problem of dictionary mutation during iteration.
These notes present the user with a problem that has no easy solution.
I dug into the implementation and found that fortunately, Antoine Pitrou already addressed this challenge (10 years ago!) by introducing an `_IterationGuard` context manager to the implementation, which delays mutation while an iteration is in progress.
I asked for confirmation and @pitrou agreed that these notes could be removed:
https://github.com/python/cpython/commit/c1baa601e2b558deb690edfdf334fceee3b03327GH-commitcomment-39514438
(cherry picked from commit 1642c0ef750f96664a98cadb09301d492098d2fb)
Co-authored-by: Daniel Fortunov <asqui@users.noreply.github.com>
Victor Stinner [Wed, 10 Jun 2020 17:24:56 +0000 (19:24 +0200)]
bpo-40826: Fix test_repl.test_close_stdin() on Windows (GH-20779) (GH-20785)
test_repl.test_close_stdin() now calls
support.suppress_msvcrt_asserts() to fix the test on Windows.
* Move suppress_msvcrt_asserts() from test.libregrtest.setup to
test.support. Make its verbose parameter optional: verbose=False by
default.
* SuppressCrashReport now uses SetErrorMode() of the msvcrt module,
rather than using ctypes.
* Remove also an unused variable (deadline) in wait_process().
Victor Stinner [Mon, 8 Jun 2020 16:44:50 +0000 (18:44 +0200)]
bpo-40910: Export Py_GetArgcArgv() function (GH-20721) (GH-20723)
Export explicitly the Py_GetArgcArgv() function to the C API and
document the function. Previously, it was exported implicitly which
no longer works since Python is built with -fvisibility=hidden.
* Add PyConfig._orig_argv member.
* Py_InitializeFromConfig() no longer calls _PyConfig_Write() twice.
* PyConfig_Read() no longer initializes Py_GetArgcArgv(): it is now
_PyConfig_Write() responsibility.
* _PyConfig_Write() result type becomes PyStatus instead of void.
* Write an unit test on Py_GetArgcArgv().
bpo-33187: Document 3.9 changes to xml.etree.ElementInclude.include (GH-20438)
Looks like the merging of bpo-33187 and bpo-20928 was racy, resulting in
this change going undocumented.
(cherry picked from commit 301f0d4ff9b6bd60599eea0612904f65a92e6dd9)
allow macOS installer builds to package pre-built html docs (GH-20715)
build-installer now looks in its directory of source tarballs
for a suitable html tarball of the same version. If so, it
will unpack and use it rather than rebuilding the html format
documentation set from the source repo. This is intended as
a speedup for test builds of the installer. Files names must
be in the same format as produced by the docs build for download,
for example, `python-3.9.0b1-docs-html.tar.bz2`.
(cherry picked from commit 63fc55b2eab0331465605a49bfd28a1bcb997f92)
Before
```
content.txt is 42 bytes in size and isa regular file.
folder is 420 bytes in size and isa directory.
magic is 4200 bytes in size and issomething else.
```
After:
```
content.txt is 42 bytes in size and is a regular file.
folder is 420 bytes in size and is a directory.
magic is 4200 bytes in size and is something else.
```
bpo-40767: Allow pure Wayland to get default XDG web browser (GH-20382)
Would be nice to backport to python 3.7+. I don't think it's worth the hassle to backport this all the way down to 3.10. But I'll let the maintainers decide.
This is hard to test because the test setup already includes this [environment variable](https://github.com/python/cpython/blob/master/Lib/test/pythoninfo.pyGH-L292)
Let me know if something doesn't match the PR guidelines. This is my first PR in the python source code.
(cherry picked from commit c822efeda9a0afe87cf3429724732fc8e19a01fb)
Co-authored-by: Jeremy Attali <jeremy.attali@gmail.com>
projects / thirdparty / Python / cpython.git / log
