fix(hooks): resolve the venv path before asking git whether it is tracked

The guard added in 9cdc40e6 was incomplete. `git ls-files` reports paths as they
are indexed and does not follow symlinks, so a repository that commits `.venv` as
a symlink to its own root alongside a tracked `bin/python` gets asked about
`.venv/bin/python` -- a path git has never heard of -- and the answer is
"untracked". The interpreter then runs. Measured on that shape: the planted
executable logged two invocations against 9cdc40e6 and none against this commit.

`repo_ships_interpreter` now resolves the bin directory with `cd -P`/`pwd -P`,
resolves the worktree root the same way, and asks git about the resolved path
relative to it. The three cases that matter all hold: a plainly committed venv is
still refused, the symlink shape is now refused, and a developer's own untracked
venv still resolves and runs.

`cd -P`/`pwd -P` rather than `realpath` or `readlink -f`, because neither is
portable to a stock macOS.
This commit is contained in:
Juan Garibay
2026-09-17 16:55:11 -04:00
parent ca1a5ad8bc
commit 3c31747016
2 changed files with 50 additions and 11 deletions
+27 -10
View File
@@ -147,6 +147,31 @@ is_pytest() {
grep -qiE 'pytest[[:space:]]+(version[[:space:]]+)?[0-9]' <<<"$version"
}
# Does the repository itself ship this interpreter?
#
# A virtualenv is never committed -- it is platform-specific binaries, and every
# Python project gitignores it. One that IS tracked is the repository handing this
# hook an executable and asking it to run. The hook is installed globally, so
# cloning a hostile repository and pushing it to your own fork would be enough,
# and on a machine with no pytest on PATH this arm is the only thing that would
# run at all. A developer's own venv is untracked, so nothing legitimate is lost.
#
# The path is resolved through symlinks before git is asked, because `git ls-files`
# reports paths as indexed and does not follow links. A repository that commits
# `.venv` as a symlink to `.` next to a tracked `bin/python` would otherwise be
# queried for `.venv/bin/python`, a path git has never heard of, and the answer
# would be "untracked". Measured: that shape ran the planted binary twice.
repo_ships_interpreter() {
local bindir real top
bindir="$(cd -P -- "$1" 2>/dev/null && pwd -P)" || return 1
[[ -n "$bindir" ]] || return 1
real="$bindir/python"
top="$(git rev-parse --show-toplevel 2>/dev/null)" || return 1
top="$(cd -P -- "$top" 2>/dev/null && pwd -P)" || return 1
[[ -n "$top" && "$real" == "$top/"* ]] || return 1
git ls-files --error-unmatch -- "${real#"$top"/}" >/dev/null 2>&1
}
resolve_pytest() {
# `${VAR+set}` rather than `-n "${VAR:-}"`, so that a variable set to nothing is
# still an override: `ECC_PYTEST_CMD=` and `ECC_PYTEST_CMD=" "` now behave
@@ -175,16 +200,8 @@ resolve_pytest() {
local venv
for venv in "${VIRTUAL_ENV:-}" .venv venv env; do
if [[ -n "$venv" && -x "$venv/bin/python" ]]; then
# A virtualenv is never committed -- it is platform-specific binaries, and
# every Python project gitignores it. One that IS tracked is the repository
# handing this hook an executable and asking it to run. The hook is installed
# globally, so cloning a hostile repository and pushing it to your own fork
# would be enough, and on a machine with no pytest on PATH this arm is the
# only thing that would run at all. Skipping costs nothing legitimate,
# because a developer's own venv is untracked -- and it says why rather than
# going quiet about it.
if git ls-files --error-unmatch -- "$venv/bin/python" >/dev/null 2>&1; then
log "Ignoring $venv/bin/python: it is tracked in this repository."
if repo_ships_interpreter "$venv/bin"; then
log "Ignoring $venv/bin/python: the repository ships it."
log " A committed virtualenv is an executable the repository controls, and"
log " this hook runs on every push in every repository."
continue
+23 -1
View File
@@ -319,6 +319,7 @@ function runHermeticPythonPrePush({
venvName = null,
venvExit = 0,
trackVenv = false,
trackedSymlinkVenv = false,
pytestCmd = null,
overrideStub = false,
pathPytestVersionLine = null,
@@ -350,6 +351,16 @@ function runHermeticPythonPrePush({
}
}
// The shape that defeats a naive `git ls-files -- .venv/bin/python` check: the
// repository commits `.venv` as a symlink to its own root plus a tracked
// `bin/python`, so git is asked about a path it has never indexed.
if (trackedSymlinkVenv) {
writeExecutable(path.join(projectDir, 'bin', 'python'), `#!/bin/sh\n${record}\nexit 0\n`);
fs.symlinkSync('.', path.join(projectDir, '.venv'));
const added = spawnSync('git', ['add', '-f', '--', 'bin/python', '.venv'], { cwd: projectDir });
assert.strictEqual(added.status, 0, added.stderr?.toString());
}
// Deliberately does NOT special-case --version: an operator's wrapper would not
// either, and the recorded calls are what prove the hook never probed it.
const overrideStubPath = overrideStub ? path.join(tempDir, 'bin', 'wrapper') : null;
@@ -416,7 +427,18 @@ if (
const { result, calls } = runHermeticPythonPrePush({ venvName: '.venv', trackVenv: true });
assert.strictEqual(result.status, 0, `${result.stdout}\n${result.stderr}`);
assert.deepStrictEqual(calls, [], JSON.stringify(calls));
assert.match(result.stdout, /it is tracked in this repository/);
assert.match(result.stdout, /the repository ships it/);
})
)
passed++;
else failed++;
if (
test('pre-push refuses a tracked interpreter reached through a committed symlink', () => {
const { result, calls } = runHermeticPythonPrePush({ trackedSymlinkVenv: true });
assert.strictEqual(result.status, 0, `${result.stdout}\n${result.stderr}`);
assert.deepStrictEqual(calls, [], JSON.stringify(calls));
assert.match(result.stdout, /the repository ships it/);
})
)
passed++;