Compare commits

...

1 Commits

Author SHA1 Message Date
Ahmed Ibrahim
66aef5f0db codex: stabilize guardian approval tests 2026-03-08 13:52:13 -07:00
6 changed files with 35 additions and 58 deletions

View File

@@ -16,6 +16,8 @@ use codex_execpolicy::RuleMatch;
use codex_protocol::models::FunctionCallOutputBody;
use codex_protocol::models::NetworkPermissions;
use codex_protocol::models::PermissionProfile;
use codex_protocol::permissions::FileSystemSandboxPolicy;
use codex_protocol::permissions::NetworkSandboxPolicy;
use codex_utils_absolute_path::AbsolutePathBuf;
use core_test_support::codex_linux_sandbox_exe_or_skip;
use core_test_support::responses::ev_assistant_message;
@@ -70,15 +72,17 @@ async fn guardian_allows_shell_additional_permissions_requests_past_policy_valid
.features
.enable(Feature::RequestPermissions)
.expect("test setup should allow enabling request permissions");
turn_context_raw
.sandbox_policy
.set(SandboxPolicy::DangerFullAccess)
.expect("test setup should allow updating sandbox policy");
// This test is about request-permissions validation, not managed sandbox
// policy enforcement. Widen the derived sandbox policies directly so the
// command runs without depending on a platform sandbox binary.
turn_context_raw.file_system_sandbox_policy =
codex_protocol::permissions::FileSystemSandboxPolicy::from(
&SandboxPolicy::DangerFullAccess,
);
FileSystemSandboxPolicy::from(turn_context_raw.sandbox_policy.get());
turn_context_raw.network_sandbox_policy =
codex_protocol::permissions::NetworkSandboxPolicy::from(&SandboxPolicy::DangerFullAccess);
NetworkSandboxPolicy::from(turn_context_raw.sandbox_policy.get());
let mut config = (*turn_context_raw.config).clone();
config.model_provider.base_url = Some(format!("{}/v1", server.uri()));
let config = Arc::new(config);
@@ -92,11 +96,14 @@ async fn guardian_allows_shell_additional_permissions_requests_past_policy_valid
turn_context_raw.provider = config.model_provider.clone();
let session = Arc::new(session);
let turn_context = Arc::new(turn_context_raw);
let expiration_ms: u64 = if cfg!(windows) { 2_500 } else { 1_000 };
let params = ExecParams {
command: if cfg!(windows) {
vec![
"cmd.exe".to_string(),
"/Q".to_string(),
"/D".to_string(),
"/C".to_string(),
"echo hi".to_string(),
]
@@ -108,7 +115,7 @@ async fn guardian_allows_shell_additional_permissions_requests_past_policy_valid
]
},
cwd: turn_context.cwd.clone(),
expiration: 1000.into(),
expiration: expiration_ms.into(),
env: HashMap::new(),
network: None,
sandbox_permissions: SandboxPermissions::WithAdditionalPermissions,

View File

@@ -90,7 +90,7 @@ fn resolve_workdir_base_path(
/// Validates feature/policy constraints for `with_additional_permissions` and
/// normalizes any path-based permissions. Errors if the request is invalid.
pub(super) fn normalize_and_validate_additional_permissions(
pub(crate) fn normalize_and_validate_additional_permissions(
request_permission_enabled: bool,
approval_policy: AskForApproval,
sandbox_permissions: SandboxPermissions,

View File

@@ -54,23 +54,26 @@ async fn test_shell_command_approval_triggers_elicitation() {
async fn shell_command_approval_triggers_elicitation() -> anyhow::Result<()> {
// Use a simple, untrusted command that creates a file so we can
// observe a side-effect.
//
// Crossplatform approach: run a tiny Python snippet to touch the file
// using `python3 -c ...` on all platforms.
let workdir_for_shell_function_call = TempDir::new()?;
let created_filename = "created_by_shell_tool.txt";
let created_file = workdir_for_shell_function_call
.path()
.join(created_filename);
let shell_command = vec![
"python3".to_string(),
"-c".to_string(),
format!("import pathlib; pathlib.Path('{created_filename}').touch()"),
];
let expected_shell_command = format_with_current_shell(&format!(
"python3 -c \"import pathlib; pathlib.Path('{created_filename}').touch()\""
));
let shell_command = if cfg!(windows) {
vec![
"New-Item".to_string(),
"-ItemType".to_string(),
"File".to_string(),
"-Path".to_string(),
created_filename.to_string(),
"-Force".to_string(),
]
} else {
vec!["touch".to_string(), created_filename.to_string()]
};
let expected_shell_command =
format_with_current_shell(&shlex::try_join(shell_command.iter().map(String::as_str))?);
let McpHandle {
process: mut mcp_process,

View File

@@ -1,18 +0,0 @@
---
source: tui/src/chatwidget/tests.rs
expression: popup
---
Experimental features
Toggle experimental features. Changes are saved to config.toml.
[ ] JavaScript REPL Enable a persistent Node-backed JavaScript REPL for interactive website debugging
and other inline JavaScript execution capabilities. Requires Node >= v22.22.0
installed.
[ ] Multi-agents Ask Codex to spawn multiple agents to parallelize the work and win in efficiency.
[ ] Apps Use a connected ChatGPT App using "$". Install Apps via /apps command. Restart
Codex after enabling.
[ ] Guardian approvals Let a guardian subagent review `on-request` approval prompts instead of showing
them to you, including sandbox escapes and blocked network access.
[ ] Prevent sleep while running Keep your computer awake while Codex is running a thread.
Press space to select or enter to save for next conversation

View File

@@ -1,19 +0,0 @@
---
source: tui/src/chatwidget/tests.rs
expression: popup
---
Experimental features
Toggle experimental features. Changes are saved to config.toml.
[ ] JavaScript REPL Enable a persistent Node-backed JavaScript REPL for interactive website debugging
and other inline JavaScript execution capabilities. Requires Node >= v22.22.0
installed.
[ ] Bubblewrap sandbox Try the new linux sandbox based on bubblewrap.
[ ] Multi-agents Ask Codex to spawn multiple agents to parallelize the work and win in efficiency.
[ ] Apps Use a connected ChatGPT App using "$". Install Apps via /apps command. Restart
Codex after enabling.
[ ] Guardian approvals Let a guardian subagent review `on-request` approval prompts instead of showing
them to you, including sandbox escapes and blocked network access.
[ ] Prevent sleep while running Keep your computer awake while Codex is running a thread.
Press space to select or enter to save for next conversation

View File

@@ -6996,10 +6996,14 @@ async fn experimental_popup_includes_guardian_approval() {
chat.open_experimental_popup();
let popup = render_bottom_popup(&chat, 120);
#[cfg(target_os = "linux")]
assert_snapshot!("experimental_popup_includes_guardian_approval_linux", popup);
#[cfg(not(target_os = "linux"))]
assert_snapshot!("experimental_popup_includes_guardian_approval", popup);
assert!(
popup.contains("Guardian approvals"),
"expected guardian approvals entry in experimental popup, got:\n{popup}"
);
assert!(
popup.contains("blocked network access"),
"expected guardian approvals description in experimental popup, got:\n{popup}"
);
}
#[tokio::test]