You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In #106045, the use case of putting a symlink in /usr/local/bin to a Python binary from another prefix was highlighted. For homebrew, this use case is used with the idea that the symlink is the public interface/location, and the prefix where it is actually installed is an implementation detail (and can change over time) (more details in astral-sh/uv#1640).
#106045 seems to have now been resolved by #127974 (I wrote a test for that). The originating issue proposed a solution which would fix the problem with the broken virtual environment in that case, but would actually break the aforementioned homebrew example by eagerly resolving the symlinked executable and using the resolved path to determine the home value in pyvenv.cfg - thereby exposing the "internal" prefix into pyvenv.cfg.
The problem I am reporting here is that
because sys._base_executable eagerly resolves symlinks in getpath.py
and venv, virtualenv and uv all use sys._base_executable to determine the home location
then when you take a virtual environment from another virtual environment in the aforementioned setup, the second virtual environment's home location will be resolved, whereas the first will not be. Thereby bleeding the implementation detail once more.
In uv, the behaviour was solidified in astral-sh/uv#8433 to avoid exposing the internal prefix in the venv. Note that virtualenv is (accidentally?) now exposing the internal prefix even on the first virtual environment pypa/virtualenv#2770.
And observe that the second venv's home is not /tmp/public/bin but the "internal detail" one.
A quick test for getpath, and a complete test for venv are included below (both failing):
diff --git a/Lib/test/test_getpath.py b/Lib/test/test_getpath.py
index f86df9d0d03..1c2e2a0b3fc 100644
--- a/Lib/test/test_getpath.py+++ b/Lib/test/test_getpath.py@@ -864,6 +864,55 @@ def test_PYTHONHOME_in_venv(self):
actual = getpath(ns, expected)
self.assertEqual(expected, actual)
+ def test_venv_w_symlinked_base_executable(self):+ """+ If we symlink the base executable, and point to it via home in pyvenv.cfg,+ we should have it as sys.executable (and sys.prefix should be the resolved location)+ """+ ns = MockPosixNamespace(+ argv0="/venv/bin/python3",+ PREFIX="/some/_internal/prefix",+ )+ # Setup venv+ ns.add_known_xfile("/venv/bin/python3")+ ns.add_known_xfile("/usr/local/bin/python3")+ ns.add_known_xfile("/some/_internal/prefix/bin/python3")++ ns.add_known_file("/venv/pyvenv.cfg", [+ # The published based executable location is /usr/local/bin - we don't want to+ # expose /some/internal/directory (this location can change under our feet)+ r"home = /usr/local/bin"+ ])+ ns.add_known_link("/venv/bin/python3", "/usr/local/bin/python3")+ ns.add_known_link("/usr/local/bin/python3", "/some/_internal/prefix/bin/python3")++ ns.add_known_file("/some/_internal/prefix/lib/python9.8/os.py")+ ns.add_known_dir("/some/_internal/prefix/lib/python9.8/lib-dynload")++ # Put a file completely outside of /usr/local to validate that the issue+ # in https://github.com/python/cpython/issues/106045 is resolved.+ ns.add_known_dir("/usr/lib/python9.8/lib-dynload")++ expected = dict(+ executable="/venv/bin/python3",+ prefix="/venv",+ exec_prefix="/venv",+ base_prefix="/some/_internal/prefix",+ base_exec_prefix="/some/_internal/prefix",+ # It is important to maintain the link to the original executable, as this+ # is used when creating a new virtual environment (which should also have home+ # set to /usr/local/bin to avoid bleeding the internal path to the venv)+ base_executable="/usr/bin/python3",+ module_search_paths_set=1,+ module_search_paths=[+ "/some/_internal/prefix/lib/python98.zip",+ "/some/_internal/prefix/lib/python9.8",+ "/some/_internal/prefix/lib/python9.8/lib-dynload",+ ],+ )+ actual = getpath(ns, expected)+ self.assertEqual(expected, actual)+
# ******************************************************************************
DEFAULT_NAMESPACE = dict(
diff --git a/Lib/test/test_venv.py b/Lib/test/test_venv.py
index 0b09010c69d..bd1b19a9c15 100644
--- a/Lib/test/test_venv.py+++ b/Lib/test/test_venv.py@@ -756,6 +756,53 @@ def test_zippath_from_non_installed_posix(self):
out, err = check_output(cmd)
self.assertTrue(zip_landmark.encode() in out)
+ @unittest.skipIf(os.name == 'nt', 'not relevant on Windows')+ @requireVenvCreate+ def test_venv_from_venv_with_symlink(self):+ """+ Test that we can create a venv from a venv using a base Python that is+ a symlink, without exposing the location of the symlink in pyvenv.cfg.+ """+ rmtree(self.env_dir)+ public_prefix = os.path.realpath(tempfile.mkdtemp())+ self.addCleanup(rmtree, public_prefix)+ public_bin_dir = os.path.join(public_prefix, 'bin')+ os.mkdir(public_bin_dir)+ public_exe = os.path.join(public_bin_dir, self.exe)+ os.symlink(sys.executable, public_exe)+ cmd = [public_exe,+ "-m",+ "venv",+ "--without-pip",+ "--without-scm-ignore-files",+ self.env_dir]++ subprocess.check_call(cmd)++ # Verify that we don't expose the internal prefix to the first venv config:+ contents = (pathlib.Path(self.env_dir) / 'pyvenv.cfg').read_text()+ self.assertIn(f'home = {public_bin_dir}\n', contents)++ # Now use the venv to make another, and assert that the internal env is+ # also not exposed there.+ second_venv = os.path.realpath(tempfile.mkdtemp())+ self.addCleanup(rmtree, second_venv)++ cmd = [os.path.join(self.env_dir, 'bin', 'python3'),+ "-m",+ "venv",+ "--without-pip",+ "--without-scm-ignore-files",+ second_venv]++ subprocess.check_call(cmd)++ contents = (pathlib.Path(second_venv) / 'pyvenv.cfg').read_text()+ self.assertIn(f'home = {public_bin_dir}\n', contents)+
@requireVenvCreate
def test_activate_shell_script_has_no_dos_newlines(self):
"""
cc @FFY00 following on from our conversation in #127974 (comment) (I would love to get a commit for the tests, if not the fix 😜).
CPython versions tested on:
CPython main branch
Operating systems tested on:
Linux
The text was updated successfully, but these errors were encountered:
There is also some useful context on cases that are covered by uv in astral-sh/uv#9723. The approach taken there allows for the additional possibility of having a symlink to Python which is not named python, and still being able to create a venv from it. The issue I've raised here is only intended to get the second and third cases right - IMO the first case should be handled by a change to the pyvenv.cfg.
Bug report
Bug description:
In #106045, the use case of putting a symlink in
/usr/local/bin
to a Python binary from another prefix was highlighted. For homebrew, this use case is used with the idea that the symlink is the public interface/location, and the prefix where it is actually installed is an implementation detail (and can change over time) (more details in astral-sh/uv#1640).#106045 seems to have now been resolved by #127974 (I wrote a test for that). The originating issue proposed a solution which would fix the problem with the broken virtual environment in that case, but would actually break the aforementioned homebrew example by eagerly resolving the symlinked executable and using the resolved path to determine the
home
value inpyvenv.cfg
- thereby exposing the "internal" prefix intopyvenv.cfg
.The problem I am reporting here is that
sys._base_executable
eagerly resolves symlinks ingetpath.py
venv
,virtualenv
anduv
all usesys._base_executable
to determine thehome
locationhome
location will be resolved, whereas the first will not be. Thereby bleeding the implementation detail once more.In
uv
, the behaviour was solidified in astral-sh/uv#8433 to avoid exposing the internal prefix in the venv. Note thatvirtualenv
is (accidentally?) now exposing the internal prefix even on the first virtual environment pypa/virtualenv#2770.To reproduce this:
And observe that the second venv's home is not
/tmp/public/bin
but the "internal detail" one.A quick test for getpath, and a complete test for venv are included below (both failing):
cc @FFY00 following on from our conversation in #127974 (comment) (I would love to get a commit for the tests, if not the fix 😜).
CPython versions tested on:
CPython main branch
Operating systems tested on:
Linux
The text was updated successfully, but these errors were encountered: