mirror of
https://github.com/openai/codex.git
synced 2026-04-24 14:45:27 +00:00
core: add host_executable integration coverage
This commit is contained in:
@@ -8,6 +8,7 @@ use codex_protocol::config_types::Settings;
|
||||
use codex_protocol::protocol::AskForApproval;
|
||||
use codex_protocol::protocol::EventMsg;
|
||||
use codex_protocol::protocol::Op;
|
||||
use codex_protocol::protocol::ReviewDecision;
|
||||
use codex_protocol::protocol::SandboxPolicy;
|
||||
use codex_protocol::user_input::UserInput;
|
||||
use core_test_support::responses::ev_assistant_message;
|
||||
@@ -19,6 +20,7 @@ use core_test_support::responses::sse;
|
||||
use core_test_support::responses::start_mock_server;
|
||||
use core_test_support::test_codex::test_codex;
|
||||
use core_test_support::wait_for_event;
|
||||
use pretty_assertions::assert_eq;
|
||||
use serde_json::Value;
|
||||
use serde_json::json;
|
||||
use std::fs;
|
||||
@@ -62,6 +64,33 @@ async fn submit_user_turn(
|
||||
Ok(())
|
||||
}
|
||||
|
||||
async fn expect_exec_approval(
|
||||
test: &core_test_support::test_codex::TestCodex,
|
||||
expected_command: &str,
|
||||
) -> codex_protocol::protocol::ExecApprovalRequestEvent {
|
||||
let event = wait_for_event(&test.codex, |event| {
|
||||
matches!(
|
||||
event,
|
||||
EventMsg::ExecApprovalRequest(_) | EventMsg::TurnComplete(_)
|
||||
)
|
||||
})
|
||||
.await;
|
||||
|
||||
match event {
|
||||
EventMsg::ExecApprovalRequest(approval) => {
|
||||
let last_arg = approval
|
||||
.command
|
||||
.last()
|
||||
.map(String::as_str)
|
||||
.unwrap_or_default();
|
||||
assert_eq!(last_arg, expected_command);
|
||||
approval
|
||||
}
|
||||
EventMsg::TurnComplete(_) => panic!("expected approval request before completion"),
|
||||
other => panic!("unexpected event: {other:?}"),
|
||||
}
|
||||
}
|
||||
|
||||
fn assert_no_matched_rules_invariant(output_item: &Value) {
|
||||
let Some(output) = output_item.get("output").and_then(Value::as_str) else {
|
||||
panic!("function_call_output should include string output payload: {output_item:?}");
|
||||
@@ -161,6 +190,192 @@ async fn execpolicy_blocks_shell_invocation() -> Result<()> {
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
#[cfg(unix)]
|
||||
async fn shell_command_absolute_path_uses_host_executable_rules() -> Result<()> {
|
||||
let git_path = "/tmp/codex-host-executable-shell/git";
|
||||
let command = format!("{git_path} status");
|
||||
|
||||
let mut builder = test_codex().with_config(move |config| {
|
||||
let policy_path = config.codex_home.join("rules").join("policy.rules");
|
||||
fs::create_dir_all(
|
||||
policy_path
|
||||
.parent()
|
||||
.expect("policy directory must have a parent"),
|
||||
)
|
||||
.expect("create policy directory");
|
||||
fs::write(
|
||||
&policy_path,
|
||||
format!(
|
||||
r#"host_executable(name = "git", paths = ["{git_path}"])
|
||||
prefix_rule(pattern = ["git", "status"], decision = "prompt")"#
|
||||
),
|
||||
)
|
||||
.expect("write policy file");
|
||||
});
|
||||
let server = start_mock_server().await;
|
||||
let test = builder.build(&server).await?;
|
||||
|
||||
let call_id = "shell-host-executable-prompt";
|
||||
let args = json!({
|
||||
"command": command,
|
||||
"timeout_ms": 1_000,
|
||||
});
|
||||
|
||||
mount_sse_once(
|
||||
&server,
|
||||
sse(vec![
|
||||
ev_response_created("resp-shell-host-exec-1"),
|
||||
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
|
||||
ev_completed("resp-shell-host-exec-1"),
|
||||
]),
|
||||
)
|
||||
.await;
|
||||
let results_mock = mount_sse_once(
|
||||
&server,
|
||||
sse(vec![
|
||||
ev_assistant_message("msg-shell-host-exec-1", "done"),
|
||||
ev_completed("resp-shell-host-exec-2"),
|
||||
]),
|
||||
)
|
||||
.await;
|
||||
|
||||
submit_user_turn(
|
||||
&test,
|
||||
"run absolute git path through shell_command",
|
||||
AskForApproval::OnRequest,
|
||||
SandboxPolicy::DangerFullAccess,
|
||||
None,
|
||||
)
|
||||
.await?;
|
||||
|
||||
let approval = expect_exec_approval(&test, &command).await;
|
||||
let reason = approval.reason.as_deref().unwrap_or_default();
|
||||
assert!(
|
||||
reason.contains("requires approval by policy"),
|
||||
"unexpected approval reason: {reason}",
|
||||
);
|
||||
|
||||
test.codex
|
||||
.submit(Op::ExecApproval {
|
||||
id: approval.effective_approval_id(),
|
||||
turn_id: None,
|
||||
decision: ReviewDecision::Denied,
|
||||
})
|
||||
.await?;
|
||||
|
||||
wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::TurnComplete(_))
|
||||
})
|
||||
.await;
|
||||
|
||||
let output_item = results_mock.single_request().function_call_output(call_id);
|
||||
let output = output_item
|
||||
.get("output")
|
||||
.and_then(Value::as_str)
|
||||
.unwrap_or_default();
|
||||
assert!(
|
||||
output.contains("rejected by user"),
|
||||
"unexpected output: {output}",
|
||||
);
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
#[cfg(unix)]
|
||||
async fn unified_exec_absolute_path_uses_host_executable_rules() -> Result<()> {
|
||||
let git_path = "/tmp/codex-host-executable-unified/git";
|
||||
let command = format!("{git_path} status");
|
||||
|
||||
let mut builder = test_codex().with_config(move |config| {
|
||||
config.features.enable(Feature::UnifiedExec);
|
||||
let policy_path = config.codex_home.join("rules").join("policy.rules");
|
||||
fs::create_dir_all(
|
||||
policy_path
|
||||
.parent()
|
||||
.expect("policy directory must have a parent"),
|
||||
)
|
||||
.expect("create policy directory");
|
||||
fs::write(
|
||||
&policy_path,
|
||||
format!(
|
||||
r#"host_executable(name = "git", paths = ["{git_path}"])
|
||||
prefix_rule(pattern = ["git", "status"], decision = "prompt")"#
|
||||
),
|
||||
)
|
||||
.expect("write policy file");
|
||||
});
|
||||
let server = start_mock_server().await;
|
||||
let test = builder.build(&server).await?;
|
||||
|
||||
let call_id = "unified-host-executable-prompt";
|
||||
let args = json!({
|
||||
"cmd": command,
|
||||
"shell": "/bin/bash",
|
||||
"yield_time_ms": 1_000,
|
||||
});
|
||||
|
||||
mount_sse_once(
|
||||
&server,
|
||||
sse(vec![
|
||||
ev_response_created("resp-unified-host-exec-1"),
|
||||
ev_function_call(call_id, "exec_command", &serde_json::to_string(&args)?),
|
||||
ev_completed("resp-unified-host-exec-1"),
|
||||
]),
|
||||
)
|
||||
.await;
|
||||
let results_mock = mount_sse_once(
|
||||
&server,
|
||||
sse(vec![
|
||||
ev_assistant_message("msg-unified-host-exec-1", "done"),
|
||||
ev_completed("resp-unified-host-exec-2"),
|
||||
]),
|
||||
)
|
||||
.await;
|
||||
|
||||
submit_user_turn(
|
||||
&test,
|
||||
"run absolute git path through exec_command",
|
||||
AskForApproval::OnRequest,
|
||||
SandboxPolicy::DangerFullAccess,
|
||||
None,
|
||||
)
|
||||
.await?;
|
||||
|
||||
let approval = expect_exec_approval(&test, &command).await;
|
||||
let reason = approval.reason.as_deref().unwrap_or_default();
|
||||
assert!(
|
||||
reason.contains("requires approval by policy"),
|
||||
"unexpected approval reason: {reason}",
|
||||
);
|
||||
|
||||
test.codex
|
||||
.submit(Op::ExecApproval {
|
||||
id: approval.effective_approval_id(),
|
||||
turn_id: None,
|
||||
decision: ReviewDecision::Denied,
|
||||
})
|
||||
.await?;
|
||||
|
||||
wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::TurnComplete(_))
|
||||
})
|
||||
.await;
|
||||
|
||||
let output_item = results_mock.single_request().function_call_output(call_id);
|
||||
let output = output_item
|
||||
.get("output")
|
||||
.and_then(Value::as_str)
|
||||
.unwrap_or_default();
|
||||
assert!(
|
||||
output.contains("rejected by user"),
|
||||
"unexpected output: {output}",
|
||||
);
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
|
||||
async fn shell_command_empty_script_with_collaboration_mode_does_not_panic() -> Result<()> {
|
||||
let server = start_mock_server().await;
|
||||
|
||||
Reference in New Issue
Block a user