To allow builds of Python to be run from an in-tree layout (rather than
an installed file layout), the VPATH variable is defined at build time
and used to locate certain landmarks - specifically,
Modules/setup.local. When this landmark is found relative to VPATH
relative to the executable, Python assumes it is running in a source
tree and generates a different default sys.path. This code remains in
release builds, so that release-ready builds can be built in-tree.
On Windows, since builds are written to 'PCbuild/', the value of
VPATH is set to '..\..', which results in a landmark of
'..\..\Modules\setup.local'. This path is outside the install directory
of Python, and may have different permissions, potentially allowing a
low-privilege user to create the landmark and an alternative `Lib`
folder that will be discovered by an otherwise restricted install.
Such a setup occurs with the legacy default install location for all
users (in the now superseded EXE installer), due to how Windows allows
all users to create folders in the root directory of their OS drive.
Our recommended mitigation on Windows is to migrate away from the
legacy installer and use the new [Python install
manager](https://www.python.org/downloads/latest/pymanager/) to install
for the current user. Installs where the directory two levels above the
Python installation directory have equivalent permissions are unaffected
(in general, a per-user install cannot be modified at all by other
users, removing any escalation of privilege risk, and could be directly
modified by a privileged user, making the potential tampering
irrelevant). Alternative mitigations might include preemptively creating
and restricting access to a `Modules` directory. Be aware that only 3.13
and 3.14 will receive updated legacy installers - earlier fixes are only
provided as sources.
Platforms other than Windows allow VPATH to be overridden, but as they
don't usually use a separated directory in the build for binaries, are
unlikely to have a landmark reference outside of the install directory.
The landmark detection involving VPATH is a fallback for when a more
specific landmark - .\pybuilddir.txt - is absent, and was included for
compatibility. Future releases of Python will no longer include the
fallback, and so builds will need to generate or preserve the
pybuilddir.txt file in order to work in-tree. This landmark file has
been generated on Windows since 3.11, and on other platforms for longer.
bz2.BZ2Decompressor objects could be reused after a decompression error. If an application caught the resulting OSError and retried with the same decompressor, crafted input could cause the decompressor to resume from an invalid internal state and perform out-of-bounds writes to a stack buffer. This could crash the process when processing untrusted data.
tarfile.data_filter could be bypassed using crafted link entries, including symlinks with empty or directory-like names, to redirect later archive members outside the intended extraction directory. This allowed a malicious tar archive to cause tarfile.extractall() to write files outside the destination directory, subject to the permissions of the extracting process.
unicodedata.normalize() can take excessive CPU time when processing
specially crafted Unicode input containing long runs of combining characters
with alternating Canonical Combining Class values.
This affects all normalization forms.
The ftpcp() function in Lib/ftplib.py was not updated when
CVE-2021-4189 was fixed. While makepasv() was patched to replace
server-supplied PASV host addresses with the actual peer address
(getpeername()[0]), ftpcp() still calls parse227() directly and passes
the raw attacker-controllable IP address and port to target.sendport(). This patch is related to CVE-2021-4189.
`xml.parsers.expat` and `xml.etree.ElementTree` use insufficient entropy for Expat hash-flooding protection, which allows a crafted XML document to trigger hash flooding.\r\n\r\nFully mitigating this vulnerability requires both updating libexpat to 2.8.0 or later and applying this patch.
If `shutil.unpack_archive()` is given a ZIP archive with an absolute Windows path containing a drive (`C:\\...`) then the archive will be extracted outside the target directory which is different than other operating systems. Only Windows is affected by this vulnerability.
http.cookies.Morsel.js_output() returns an inline <script> snippet and only escapes " for JavaScript string context. It does not neutralize the HTML parser-sensitive sequence </script> inside the generated script element. Mitigation base64-encodes the cookie value to disallow escaping using cookie value.
The method "sock_recvfrom_into()" of "asyncio.ProacterEventLoop" (Windows only) was missing a boundary check for the data buffer when using nbytes parameter. This allowed for an out-of-bounds buffer write if data was larger than the buffer size. Non-Windows platforms are not affected.
The "profiling.sampling" module (Python 3.15+) and "asyncio introspection capabilities" (3.14+, "python -m asyncio ps" and "python -m asyncio pstree") features could be used to read and write addresses in a privileged process if that process connected to a malicious or "infected" Python process via the remote debugging feature. This vulnerability requires persistently and repeatedly connecting to the process to be exploited, even after the connecting process crashes with high likelihood due to ASLR.
Mitgation of CVE-2026-4519 was incomplete. If the URL contained "%action" the mitigation could be bypassed for certain browser types the "webbrowser.open()" API could have commands injected into the underlying shell. See CVE-2026-4519 for details.
Use-after-free (UAF) was possible in the `lzma.LZMADecompressor`, `bz2.BZ2Decompressor`, and `gzip.GzipFile` when a memory allocation fails with a `MemoryError` and the decompression instance is re-used. This scenario can be triggered if the process is under memory pressure. The fix cleans up the dangling pointer in this specific error condition.
The vulnerability is only present if the program re-uses decompressor instances across multiple decompression calls even after a `MemoryError` is raised during decompression. Using the helper functions to one-shot decompress data such as `lzma.decompress()`, `bz2.decompress()`, `gzip.decompress()`, and `zlib.decompress()` are not affected as a new decompressor instance is used per call. If the decompressor instance is not re-used after an error condition, this usage is similarly not vulnerable.
When calling base64.b64decode() or related functions the decoding process would stop after encountering the first padded quad regardless of whether there was more information to be processed. This can lead to data being accepted which may be processed differently by other implementations. Use "validate=True" to enable stricter processing of base64 data.
The webbrowser.open() API would accept leading dashes in the URL which
could be handled as command line options for certain web browsers. New
behavior rejects leading dashes. Users are recommended to sanitize URLs
prior to passing to webbrowser.open().
When an Expat parser with a registered ElementDeclHandler parses an inline
document type definition containing a deeply nested content model a C stack
overflow occurs.
The fix for CVE-2026-0672, which rejected control characters in http.cookies.Morsel, was incomplete. The Morsel.update(), |= operator, and unpickling paths were not patched, allowing control characters to bypass input validation. Additionally, BaseCookie.js_output() lacked the output validation applied to BaseCookie.output().
The "tarfile" module would still apply normalization of AREGTYPE (\x00) blocks to DIRTYPE, even while processing a multi-block member such as GNUTYPE_LONGNAME or GNUTYPE_LONGLINK. This could result in a crafted tar archive being misinterpreted by the tarfile module compared to other implementations.
The import hook in CPython that handles legacy *.pyc files (SourcelessFileLoader) is incorrectly handled in FileLoader (a base class) and so does not use io.open_code() to read the .pyc files. sys.audit handlers for this audit event therefore do not fire.
The
email module, specifically the "BytesGenerator" class, didn’t properly quote newlines for email headers when
serializing an email message allowing for header injection when an email
is serialized. This is only applicable if using "LiteralHeader" writing headers that don't respect email folding rules, the new behavior will reject the incorrectly folded headers in "BytesGenerator".
When passing data to the b64decode(), standard_b64decode(), and urlsafe_b64decode() functions in the "base64" module the characters "+/" will always be accepted, regardless of the value of "altchars" parameter, typically used to establish an "alternative base64 alphabet" such as the URL safe alphabet. This behavior matches what is recommended in earlier base64 RFCs, but newer RFCs now recommend either dropping characters outside the specified base64 alphabet or raising an error. The old behavior has the possibility of causing data integrity issues. This behavior can only be insecure if your application uses an alternate base64 alphabet (without "+/"). If your application does not use the "altchars" parameter or the urlsafe_b64decode() function, then your application does not use an alternative base64 alphabet. The attached patches DOES NOT make the base64-decode behavior raise an error, as this would be a change in behavior and break existing programs. Instead, the patch deprecates the behavior which will be replaced with the newly recommended behavior in a future version of Python. Users are recommended to mitigate by verifying user-controlled inputs match the base64 alphabet they are expecting or verify that their application would not be affected if the b64decode() functions accepted "+" or "/" outside of altchars.