Skip to content

OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities. npm: https://www.npmjs.com/package/@openai/codex-security

fix(scans): archive long POSIX output directory names (#2529)

Michael D'Angelo3254a2159242
diff --git a/plugins/codex-security/scripts/workbench/storage.py b/plugins/codex-security/scripts/workbench/storage.py
index 85e5933f2..28a31a947 100644
--- a/plugins/codex-security/scripts/workbench/storage.py
+++ b/plugins/codex-security/scripts/workbench/storage.py
@@ -2,6 +2,7 @@
from __future__ import annotations
+import hashlib
import os
import sqlite3
import tempfile
@@ -35,18 +36,37 @@ def worker_usage_directory(scan_id: str) -> Path:
return state_dir() / "codex-home" / "worker-usage" / scan_id
+def short_archive_prefix(original_name: str) -> str:
+ """Bind a short archive name to the complete original filename."""
+ return f".previous-{hashlib.sha256(os.fsencode(original_name)).hexdigest()}-"
+
+
+def archive_source_directory(archive: Path, source: Path) -> Path | None:
+ """Return the original sibling when the archived name identifies it."""
+ if not source.is_relative_to(archive.parent):
+ return None
+ parts = source.relative_to(archive.parent).parts
+ if not parts or ".." in parts:
+ return None
+ name = parts[0]
+ if archive.name.startswith(f"{name}.previous-") or (
+ archive.name.startswith(".previous-")
+ and archive.name.startswith(short_archive_prefix(name))
+ ):
+ return archive.parent / name
+ return None
+
+
def archived_deep_path(scan_dir: Path, value: str | None) -> str | None:
"""Resolve old archive rows that still name the original sibling output."""
if value is None or ".previous-" not in scan_dir.name:
return value
path = Path(value)
- if not path.is_relative_to(scan_dir.parent):
- return value
- parts = path.relative_to(scan_dir.parent).parts
- if not parts or ".." in parts or not scan_dir.name.startswith(f"{parts[0]}.previous-"):
+ original = archive_source_directory(scan_dir, path)
+ if original is None:
return value
# Keep missing files inside the archive too. Never follow the replacement scan's output.
- return str(scan_dir.joinpath(*parts[1:]))
+ return str(scan_dir / path.relative_to(original))
def saved_deep_workers(connection: sqlite3.Connection, scan: sqlite3.Row) -> list[Any]:
diff --git a/plugins/codex-security/scripts/workbench_scan_start.py b/plugins/codex-security/scripts/workbench_scan_start.py
index de93d0429..075d9ef38 100644
--- a/plugins/codex-security/scripts/workbench_scan_start.py
+++ b/plugins/codex-security/scripts/workbench_scan_start.py
@@ -3,6 +3,7 @@
from __future__ import annotations
import argparse
+import errno
import json
import os
import re
@@ -19,7 +20,12 @@ from typing import Any
sys.path.insert(0, str(Path(__file__).resolve().parent))
from filesystem_identity import serialize_filesystem_identity
from finalize_scan_contract import write_scan_local_bytes
-from workbench.storage import create_private_temporary_directory, state_dir
+from workbench.storage import (
+ archive_source_directory,
+ create_private_temporary_directory,
+ short_archive_prefix,
+ state_dir,
+)
from workbench_feedback import get_scan_feedback
from workbench_target import (
directory_content_digest,
@@ -129,7 +135,7 @@ def archive_scan(
if archived_scan_dir is not None and (
not args.archive_existing
or archived_scan_dir.parent != scan_dir.parent
- or not archived_scan_dir.name.startswith(f"{scan_dir.name}.previous-")
+ or archive_source_directory(archived_scan_dir, scan_dir) != scan_dir
):
raise SystemExit("The archived scan must be a previous sibling of the scan directory.")
@@ -179,9 +185,17 @@ def archive_scan(
)
if before_archive is not None:
before_archive()
- archived_scan_dir = Path(
- tempfile.mkdtemp(prefix=f"{scan_dir.name}.previous-", dir=scan_dir.parent)
- ).resolve()
+ try:
+ reserved = tempfile.mkdtemp(
+ prefix=f"{scan_dir.name}.previous-", dir=scan_dir.parent
+ )
+ except OSError as error:
+ if os.name != "posix" or error.errno != errno.ENAMETOOLONG:
+ raise
+ reserved = tempfile.mkdtemp(
+ prefix=short_archive_prefix(scan_dir.name), dir=scan_dir.parent
+ )
+ archived_scan_dir = Path(reserved).resolve()
archived_scan_dir.rmdir()
scan_dir.rename(archived_scan_dir)
moved = True
diff --git a/plugins/codex-security/scripts/workbench_scan_usage.py b/plugins/codex-security/scripts/workbench_scan_usage.py
index dc4b7a845..5af54fef4 100644
--- a/plugins/codex-security/scripts/workbench_scan_usage.py
+++ b/plugins/codex-security/scripts/workbench_scan_usage.py
@@ -20,7 +20,7 @@ from typing import Any, Mapping
from urllib.parse import quote
sys.path.insert(0, str(Path(__file__).resolve().parent))
-from workbench.storage import worker_usage_directory
+from workbench.storage import archive_source_directory, worker_usage_directory
TOKEN_FIELDS = {
"input_tokens": "inputTokens",
@@ -708,10 +708,12 @@ def _discover_rollout_sessions(
for path in worker_directories
if Path(path).is_relative_to(scan_directory)
}
- for marker in re.finditer(r"\.previous-", scan_directory.name):
- if marker.start() == 0:
- continue
- original = scan_directory.with_name(scan_directory.name[: marker.start()])
+ originals = {
+ original
+ for path in bound_directories | {Path(path) for path in worker_directories}
+ if (original := archive_source_directory(scan_directory, path)) is not None
+ }
+ for original in originals:
candidates = {original / relative: path for relative, path in relatives.items()}
# Older archives kept absolute discovery paths in their original directory.
if (
diff --git a/plugins/codex-security/tests/test_workbench_scan_archive.py b/plugins/codex-security/tests/test_workbench_scan_archive.py
index bc6cd3fd7..93d94f552 100644
--- a/plugins/codex-security/tests/test_workbench_scan_archive.py
+++ b/plugins/codex-security/tests/test_workbench_scan_archive.py
@@ -160,6 +160,46 @@ def test_archive_requires_previous_directory_when_artifacts_exist(tmp_path: Path
assert "archived scan directory is required" in str(rejected["stderr"])
[email protected](os.name != "posix", reason="uses the filesystem's POSIX component limit")
+def test_archive_keeps_output_with_a_full_length_filename(tmp_path: Path) -> None:
+ limit = os.pathconf(tmp_path, "PC_NAME_MAX")
+ if limit < 0:
+ pytest.skip("This filesystem does not report a component limit")
+ name = "a" * (limit - 1) + "b"
+ state_dir = tmp_path / "state"
+ repository = tmp_path / "repository"
+ scan_dir = tmp_path / name
+ repository.mkdir()
+ scan_dir.mkdir(mode=0o700)
+ previous = register_scan(state_dir, repository, scan_dir)
+ fail_scan(state_dir, str(previous["scanId"]), "Interrupted fixture scan.")
+ (scan_dir / "report.md").write_text("original report\n")
+
+ current = register_scan(state_dir, repository, scan_dir, "--archive-existing")
+ archived = Path(str(current["archivedScanDir"]))
+ assert archived.parent == tmp_path
+ assert (archived / "report.md").read_text() == "original report\n"
+ assert not (scan_dir / "report.md").exists()
+ with sqlite3.connect(state_dir / "workbench.sqlite3") as connection:
+ assert connection.execute(
+ "SELECT scan_dir FROM scans WHERE id = ?", (previous["scanId"],)
+ ).fetchone() == (str(archived),)
+
+ other = tmp_path / (name[:-1] + "c")
+ other.mkdir(mode=0o700)
+ rejected = register_scan(
+ state_dir,
+ repository,
+ other,
+ "--archive-existing",
+ "--archived-scan-dir",
+ str(archived),
+ check=False,
+ )
+ assert rejected["returncode"] != 0
+ assert "previous sibling" in str(rejected["stderr"])
+
+
def test_archive_does_not_replace_a_running_scan(tmp_path: Path) -> None:
state_dir = tmp_path / "state"
repository = tmp_path / "repository"
diff --git a/plugins/codex-security/tests/test_workbench_scan_usage.py b/plugins/codex-security/tests/test_workbench_scan_usage.py
index b20457ad4..3e51d5f2a 100644
--- a/plugins/codex-security/tests/test_workbench_scan_usage.py
+++ b/plugins/codex-security/tests/test_workbench_scan_usage.py
@@ -2,6 +2,7 @@
from __future__ import annotations
+import hashlib
import json
import sqlite3
import subprocess
@@ -3158,11 +3159,21 @@ def test_cancel_scan_retains_runtime_usage_on_repeated_cancellation(tmp_path: Pa
@pytest.mark.parametrize(
- ("worker_kind", "legacy_archive"),
- [("discovery", False), ("discovery", True), ("dedup", False)],
+ ("worker_kind", "legacy_archive", "short_archive"),
+ [
+ ("discovery", False, False),
+ ("discovery", True, False),
+ ("dedup", False, False),
+ pytest.param(
+ "discovery",
+ True,
+ True,
+ marks=pytest.mark.skipif(sys.platform == "win32", reason="POSIX archive naming"),
+ ),
+ ],
)
def test_completion_recovers_replaced_deep_worker_threads(
- tmp_path: Path, worker_kind: str, legacy_archive: bool
+ tmp_path: Path, worker_kind: str, legacy_archive: bool, short_archive: bool
) -> None:
# Compressed variants exercise the bundled Node decoder in the SDK runtime lane.
fixture = _start_scan(tmp_path, mode="deep")
@@ -3307,6 +3318,13 @@ def test_completion_recovers_replaced_deep_worker_threads(
"threadCount": 5,
}.items()
)
+ archived_arguments = []
+ if short_archive:
+ digest = hashlib.sha256(fixture.scan_dir.name.encode()).hexdigest()
+ archived = fixture.scan_dir.with_name(f".previous-{digest}-fixture")
+ fixture.scan_dir.rename(archived)
+ fixture.scan_dir.mkdir(mode=0o700)
+ archived_arguments = ["--archived-scan-dir", str(archived)]
run_workbench(
fixture.state_dir,
"register-cli-scan",
@@ -3315,6 +3333,7 @@ def test_completion_recovers_replaced_deep_worker_threads(
"--scan-dir",
str(fixture.scan_dir),
"--archive-existing",
+ *archived_arguments,
"--recipe-json",
json.dumps(
{
diff --git a/sdk/typescript/docs/cli.md b/sdk/typescript/docs/cli.md
index 4de4634e0..a7d650b0d 100644
--- a/sdk/typescript/docs/cli.md
+++ b/sdk/typescript/docs/cli.md
@@ -538,7 +538,9 @@ through inheritance. Other users must not have permission to set reparse data
on the destination directory itself.
`--archive-existing` moves earlier output to a sibling directory with a unique
suffix, such as `results.previous-ab12cd34` for an output directory named
-`results`. Use `--dry-run` to preview the move.
+`results`. If adding the suffix exceeds the filesystem's filename limit on POSIX,
+the archive uses a short sibling name starting with `.previous-` and a hash of the
+original filename. Use `--dry-run` to preview the move.
Scans also attempt to write SARIF to `<scan-dir>/exports/results.sarif`. If that
optional export fails, the scan can still complete. Use the
diff --git a/sdk/typescript/scripts/package-dist-files.mjs b/sdk/typescript/scripts/package-dist-files.mjs
index 6b7e809e8..c3cefbaf7 100644
--- a/sdk/typescript/scripts/package-dist-files.mjs
+++ b/sdk/typescript/scripts/package-dist-files.mjs
@@ -133,6 +133,7 @@ const distModules = [
"retry-windows-rename",
"runtime",
"scan-activity",
+ "scan-archive-path",
"scan-comparison",
"scan-dashboard",
"scan-history-renderer",
diff --git a/sdk/typescript/src/runtime.ts b/sdk/typescript/src/runtime.ts
index 1c4aa5717..1b05ab886 100644
--- a/sdk/typescript/src/runtime.ts
+++ b/sdk/typescript/src/runtime.ts
@@ -16,6 +16,7 @@ import {
type SpawnOptionsWithoutStdio,
} from "node:child_process";
import { randomUUID } from "node:crypto";
+import { shortArchivePrefix } from "./scan-archive-path.js";
import {
chmodSync,
close as closeCallback,
@@ -2173,14 +2174,33 @@ export async function planOutputArchive(
if (outputDirectory === null) return null;
const entries = await readdir(outputDirectory).catch(nullIfMissingFileError);
if (entries === null || entries.length === 0) return null;
- return `${outputDirectory}${outputArchiveSuffix()}`;
+ const id = randomUUID().slice(0, 8);
+ const archive = `${outputDirectory}${outputArchiveSuffix(id)}`;
+ if (process.platform === "win32" || !(await archiveNameTooLong(archive)))
+ return archive;
+ // The workbench reserves a sibling with an eight-character random suffix.
+ const sibling = `${outputDirectory}.previous-${id}`;
+ if (!(await archiveNameTooLong(sibling))) return sibling;
+ return join(
+ dirname(outputDirectory),
+ shortArchivePrefix(basename(outputDirectory)) + id,
+ );
+}
+
+async function archiveNameTooLong(path: string): Promise<boolean> {
+ try {
+ await lstat(path);
+ return false;
+ } catch (error) {
+ return nodeErrorCode(error) === "ENAMETOOLONG";
+ }
}
-function outputArchiveSuffix(): string {
+function outputArchiveSuffix(id = randomUUID().slice(0, 8)): string {
return `.previous-${new Date()
.toISOString()
.replaceAll(/[-:]/g, "")
- .replace(/\.\d{3}Z$/, "")}-${randomUUID().slice(0, 8)}`;
+ .replace(/\.\d{3}Z$/, "")}-${id}`;
}
export function requireModelSafeOutputDir(path: string): void {
diff --git a/sdk/typescript/src/scan-archive-path.ts b/sdk/typescript/src/scan-archive-path.ts
new file mode 100644
index 000000000..e093b2126
--- /dev/null
+++ b/sdk/typescript/src/scan-archive-path.ts
@@ -0,0 +1,5 @@
+import { createHash } from "node:crypto";
+
+export function shortArchivePrefix(originalName: string): string {
+ return `.previous-${createHash("sha256").update(originalName).digest("hex")}-`;
+}
diff --git a/sdk/typescript/src/scan-logs.ts b/sdk/typescript/src/scan-logs.ts
index c83a3fc3a..1fa72cc03 100644
--- a/sdk/typescript/src/scan-logs.ts
+++ b/sdk/typescript/src/scan-logs.ts
@@ -1,4 +1,4 @@
-import { basename, dirname, join, relative } from "node:path";
+import { basename, dirname, isAbsolute, join, relative, sep } from "node:path";
import { isDeepStrictEqual } from "node:util";
import { firstNativeSessionLine, sessionLines } from "./session-lines.js";
import { parseJson } from "./value.js";
@@ -13,6 +13,7 @@ import {
sessionStartedAt,
} from "./scan-sessions.js";
import { compareTimestamps } from "./timestamps.js";
+import { shortArchivePrefix } from "./scan-archive-path.js";
interface ScanLogOptions {
scanId: string;
@@ -164,12 +165,34 @@ export async function readScanLogs(options: ScanLogOptions) {
if (options.scanDirectory && startedAt !== null) {
const directories = [options.scanDirectory];
const name = basename(options.scanDirectory);
+ const parent = dirname(options.scanDirectory);
+ const originals = new Set<string>();
for (const marker of name.matchAll(/\.previous-/g)) {
- if (!marker.index) continue;
- const original = join(
- dirname(options.scanDirectory),
- name.slice(0, marker.index),
- );
+ if (marker.index)
+ originals.add(join(parent, name.slice(0, marker.index)));
+ }
+ if (name.startsWith(".previous-")) {
+ // A short archive has no readable source name. Only already-bound
+ // sessions can supply a sibling whose complete filename matches its hash.
+ const boundCwds = [
+ root?.workingDirectory,
+ ...Array.from(traversed, (id) => logs.get(id)?.[0].workingDirectory),
+ ];
+ for (const cwd of boundCwds) {
+ if (!cwd) continue;
+ const path = relative(parent, cwd);
+ const sibling = path.split(sep, 1)[0];
+ if (
+ sibling &&
+ sibling !== ".." &&
+ !isAbsolute(path) &&
+ name.startsWith(shortArchivePrefix(sibling))
+ ) {
+ originals.add(join(parent, sibling));
+ }
+ }
+ }
+ for (const original of originals) {
// A Desktop owner can work outside the scan directory. Its registered
// workers provide the same binding after archiving the output.
if (
diff --git a/sdk/typescript/tests-ts/runtime.test.ts b/sdk/typescript/tests-ts/runtime.test.ts
index 9ddcd220b..d46e8f91c 100644
--- a/sdk/typescript/tests-ts/runtime.test.ts
+++ b/sdk/typescript/tests-ts/runtime.test.ts
@@ -6405,6 +6405,31 @@ describe("runtime directories and plugin Python boundary", () => {
).not.toThrow();
});
+ testPosix.each([
+ [230, "a".repeat(230)],
+ [250, "é".repeat(125)],
+ ] as const)(
+ "previews and archives a %i-byte output name",
+ async (_bytes, name) => {
+ const root = await temporaryDirectory();
+ const output = join(root, name);
+ await mkdir(output, { mode: 0o700 });
+ await writeFile(join(output, "report.md"), "first report");
+ const preview = await planOutputArchive(output);
+ expect(dirname(preview!)).toBe(root);
+ await expect(lstat(preview!)).rejects.toMatchObject({ code: "ENOENT" });
+ const archived: string[] = [];
+ await prepareOutputDir(output, "repo", root, undefined, true, (path) => {
+ archived.push(path);
+ });
+ expect(archived).toHaveLength(1);
+ expect(await readFile(join(archived[0]!, "report.md"), "utf8")).toBe(
+ "first report",
+ );
+ expect(await readdir(output)).toEqual([]);
+ },
+ );
+
test("prepares an output directory without moving prior results", async () => {
const root = await temporaryDirectory();
const output = join(root, "scan");
diff --git a/sdk/typescript/tests-ts/scan-logs.test.ts b/sdk/typescript/tests-ts/scan-logs.test.ts
index cd5bed3f4..f0b69c68d 100644
--- a/sdk/typescript/tests-ts/scan-logs.test.ts
+++ b/sdk/typescript/tests-ts/scan-logs.test.ts
@@ -1,4 +1,5 @@
import { writeJsonLines } from "./support/json.js";
+import { createHash } from "node:crypto";
import { mkdir, readFile, rename, writeFile } from "node:fs/promises";
import { unlinkSync } from "node:fs";
import { join } from "node:path";
@@ -173,9 +174,14 @@ describe("saved scan logs", () => {
expect(result.sessions.map(({ threadId }) => threadId)).toEqual(["parent"]);
});
- test.each([false, true])(
- "collects parentless comparisons inside a shared Desktop scan window (archived: %p)",
- async (archived) => {
+ test.each([
+ "original",
+ "legacy archive",
+ "short archive",
+ "unrelated short archive",
+ ])(
+ "collects parentless comparisons inside a shared Desktop scan window (%s)",
+ async (layout) => {
const home = await temporaryHome();
const scanDir = join(home, "scan");
const cwd = join(scanDir, "artifacts", "deep_discovery");
@@ -213,10 +219,27 @@ describe("saved scan logs", () => {
"2026-08-02T01:00:00Z",
cwd,
);
+ await writeSession(
+ home,
+ "unbound-sibling",
+ [],
+ undefined,
+ "2026-08-02T01:00:00Z",
+ join(home, "scan-other", "artifacts"),
+ );
+ const digest = createHash("sha256")
+ .update(layout === "unrelated short archive" ? "scan-other" : "scan")
+ .digest("hex");
+ const directory =
+ layout === "original"
+ ? scanDir
+ : layout === "legacy archive"
+ ? scanDir + ".previous-test"
+ : join(home, `.previous-${digest}-test`);
const result = await readSavedScanLogs(
{
scanId: "scan-1",
- scanDir: archived ? scanDir + ".previous-test" : scanDir,
+ scanDir: directory,
mode: "deep",
continuationThreadId: "owner",
lifecycle: {
@@ -228,11 +251,11 @@ describe("saved scan logs", () => {
},
home,
);
- expect(result.sessions.map(({ threadId }) => threadId).sort()).toEqual([
- "comparison",
- "owner",
- "worker",
- ]);
+ expect(result.sessions.map(({ threadId }) => threadId).sort()).toEqual(
+ layout === "unrelated short archive"
+ ? ["owner", "worker"]
+ : ["comparison", "owner", "worker"],
+ );
},
);
@@ -832,6 +855,11 @@ describe("saved scan logs", () => {
]);
expect(JSON.stringify(archivedLogs)).toContain("review archived scan");
expect(JSON.stringify(archivedLogs)).not.toContain("PRIVATE REPLACEMENT");
+ const digest = createHash("sha256").update("results").digest("hex");
+ const shortArchive = join(home, "scans", `.previous-${digest}-a1b2c3d4`);
+ expect(
+ await readScanLogs({ ...options, scanDirectory: shortArchive }),
+ ).toEqual(archivedLogs);
for (const logCompletedAt of [null, undefined]) {
const legacyLogs = await readSavedScanLogs(