openai/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-11-02 19:43:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge 7a13c6a1b6 into sapling-pr-archive-bolinfest

Browse Source
This commit is contained in:
Michael Bolin
2025-11-21 09:50:14 -08:00
committed by GitHub
parent 80c8dd0588 7a13c6a1b6
commit 40a4579995
48 changed files with 587 additions and 360 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/issue-deduplicator.yml
-1
View File
@@ -46,7 +46,6 @@ jobs:
with:
openai-api-key: ${{ secrets.CODEX_OPENAI_API_KEY }}
allow-users: "*"
model: gpt-5.1
prompt: |
You are an assistant that triages new GitHub issues by identifying potential duplicates.
.github/workflows/rust-release.yml
+8
View File
@@ -407,6 +407,14 @@ jobs:
- name: List
run: ls -R dist/
# This is a temporary fix: we should modify shell-tool-mcp.yml so these
# files do not end up in dist/ in the first place.
- name: Delete entries from dist/ that should not go in the release
run: |
rm -rf dist/shell-tool-mcp*
ls -R dist/
- name: Define release name
id: release_name
run: |
codex-rs/Cargo.lock
Generated
+6 -1
View File
@@ -187,8 +187,10 @@ dependencies = [
"codex-app-server-protocol",
"codex-core",
"codex-protocol",
"core_test_support",
"serde",
"serde_json",
"shlex",
"tokio",
"uuid",
"wiremock",
@@ -860,7 +862,6 @@ dependencies = [
"codex-login",
"codex-protocol",
"codex-utils-json-to-toml",
"codex-windows-sandbox",
"core_test_support",
"mcp-types",
"opentelemetry-appender-tracing",
@@ -869,6 +870,7 @@ dependencies = [
"serde",
"serde_json",
"serial_test",
"shlex",
"tempfile",
"tokio",
"toml",
@@ -1777,6 +1779,7 @@ dependencies = [
"notify",
"regex-lite",
"serde_json",
"shlex",
"tempfile",
"tokio",
"walkdir",
@@ -3738,11 +3741,13 @@ dependencies = [
"assert_cmd",
"codex-core",
"codex-mcp-server",
"core_test_support",
"mcp-types",
"os_info",
"pretty_assertions",
"serde",
"serde_json",
"shlex",
"tokio",
"wiremock",
]
codex-rs/Cargo.toml
-1
View File
@@ -169,7 +169,6 @@ reqwest = "0.12"
rmcp = { version = "0.8.5", default-features = false }
schemars = "0.8.22"
seccompiler = "0.5.0"
sentry = "0.34.0"
serde = "1"
serde_json = "1"
serde_with = "3.14"
codex-rs/app-server-protocol/src/protocol/v2.rs
+5 -2
View File
@@ -50,6 +50,9 @@ macro_rules! v2_enum_from_core {
}
/// This translation layer make sure that we expose codex error code in camel case.
///
/// When an upstream HTTP status is available (for example, from the Responses API or a provider),
/// it is forwarded in `httpStatusCode` on the relevant `codexErrorInfo` variant.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -615,7 +618,7 @@ pub struct Turn {
#[error("{message}")]
pub struct TurnError {
pub message: String,
pub codex_error_code: Option<CodexErrorInfo>,
pub codex_error_info: Option<CodexErrorInfo>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1253,7 +1256,7 @@ mod tests {
}
#[test]
fn codex_error_code_serializes_http_status_code_in_camel_case() {
fn codex_error_info_serializes_http_status_code_in_camel_case() {
let value = CodexErrorInfo::ResponseTooManyFailedAttempts {
http_status_code: Some(401),
};
codex-rs/app-server/Cargo.toml
+1 -1
View File
@@ -40,7 +40,6 @@ tracing = { workspace = true, features = ["log"] }
tracing-subscriber = { workspace = true, features = ["env-filter", "fmt"] }
opentelemetry-appender-tracing = { workspace = true }
uuid = { workspace = true, features = ["serde", "v7"] }
codex-windows-sandbox.workspace = true
[dev-dependencies]
app_test_support = { workspace = true }
@@ -54,3 +53,4 @@ serial_test = { workspace = true }
tempfile = { workspace = true }
toml = { workspace = true }
wiremock = { workspace = true }
shlex = { workspace = true }
codex-rs/app-server/README.md
+23
View File
@@ -339,6 +339,29 @@ Event notifications are the server-initiated event stream for thread lifecycles,
The app-server streams JSON-RPC notifications while a turn is running. Each turn starts with `turn/started` (initial `turn`) and ends with `turn/completed` (final `turn` plus token `usage`), and clients subscribe to the events they care about, rendering each item incrementally as updates arrive. The per-item lifecycle is always: `item/started` → zero or more item-specific deltas → `item/completed`.
- `turn/started` — `{ turn }` with the turn id, empty `items`, and `status: "inProgress"`.
- `turn/completed` — `{ turn }` where `turn.status` is `completed`, `interrupted`, or `failed`; failures carry `{ error: { message, codexErrorInfo? } }`.
Today both notifications carry an empty `items` array even when item events were streamed; rely on `item/*` notifications for the canonical item list until this is fixed.
#### Errors
`error` event is emitted whenever the server hits an error mid-turn (for example, upstream model errors or quota limits). Carries the same `{ error: { message, codexErrorInfo? } }` payload as `turn.status: "failed"` and may precede that terminal notification.
`codexErrorInfo` maps to the `CodexErrorInfo` enum. Common values:
- `ContextWindowExceeded`
- `UsageLimitExceeded`
- `HttpConnectionFailed { httpStatusCode? }`: upstream HTTP failures including 4xx/5xx
- `ResponseStreamConnectionFailed { httpStatusCode? }`: failure to connect to the response SSE stream
- `ResponseStreamDisconnected { httpStatusCode? }`: disconnect of the response SSE stream in the middle of a turn before completion
- `ResponseTooManyFailedAttempts { httpStatusCode? }`
- `BadRequest`
- `Unauthorized`
- `SandboxError`
- `InternalServerError`
- `Other`: all unclassified errors
When an upstream HTTP status is available (for example, from the Responses API or a provider), it is forwarded in `httpStatusCode` on the relevant `codexErrorInfo` variant.
#### Thread items
`ThreadItem` is the tagged union carried in turn responses and `item/*` notifications. Currently we support events for the following items:
codex-rs/app-server/src/bespoke_event_handling.rs
+11 -11
View File
@@ -264,7 +264,7 @@ pub(crate) async fn apply_bespoke_event_handling(
EventMsg::Error(ev) => {
let turn_error = TurnError {
message: ev.message,
codex_error_code: ev.codex_error_code.map(V2CodexErrorInfo::from),
codex_error_info: ev.codex_error_info.map(V2CodexErrorInfo::from),
};
handle_error(conversation_id, turn_error.clone(), &turn_summary_store).await;
outgoing
@@ -278,7 +278,7 @@ pub(crate) async fn apply_bespoke_event_handling(
// but we notify the client.
let turn_error = TurnError {
message: ev.message,
codex_error_code: ev.codex_error_code.map(V2CodexErrorInfo::from),
codex_error_info: ev.codex_error_info.map(V2CodexErrorInfo::from),
};
outgoing
.send_server_notification(ServerNotification::Error(ErrorNotification {
@@ -899,7 +899,7 @@ mod tests {
conversation_id,
TurnError {
message: "boom".to_string(),
codex_error_code: Some(V2CodexErrorInfo::InternalServerError),
codex_error_info: Some(V2CodexErrorInfo::InternalServerError),
},
&turn_summary_store,
)
@@ -910,7 +910,7 @@ mod tests {
turn_summary.last_error,
Some(TurnError {
message: "boom".to_string(),
codex_error_code: Some(V2CodexErrorInfo::InternalServerError),
codex_error_info: Some(V2CodexErrorInfo::InternalServerError),
})
);
Ok(())
@@ -956,7 +956,7 @@ mod tests {
conversation_id,
TurnError {
message: "oops".to_string(),
codex_error_code: None,
codex_error_info: None,
},
&turn_summary_store,
)
@@ -996,7 +996,7 @@ mod tests {
conversation_id,
TurnError {
message: "bad".to_string(),
codex_error_code: Some(V2CodexErrorInfo::Other),
codex_error_info: Some(V2CodexErrorInfo::Other),
},
&turn_summary_store,
)
@@ -1024,7 +1024,7 @@ mod tests {
TurnStatus::Failed {
error: TurnError {
message: "bad".to_string(),
codex_error_code: Some(V2CodexErrorInfo::Other),
codex_error_info: Some(V2CodexErrorInfo::Other),
}
}
);
@@ -1079,7 +1079,7 @@ mod tests {
conversation_a,
TurnError {
message: "a1".to_string(),
codex_error_code: Some(V2CodexErrorInfo::BadRequest),
codex_error_info: Some(V2CodexErrorInfo::BadRequest),
},
&turn_summary_store,
)
@@ -1098,7 +1098,7 @@ mod tests {
conversation_b,
TurnError {
message: "b1".to_string(),
codex_error_code: None,
codex_error_info: None,
},
&turn_summary_store,
)
@@ -1134,7 +1134,7 @@ mod tests {
TurnStatus::Failed {
error: TurnError {
message: "a1".to_string(),
codex_error_code: Some(V2CodexErrorInfo::BadRequest),
codex_error_info: Some(V2CodexErrorInfo::BadRequest),
}
}
);
@@ -1155,7 +1155,7 @@ mod tests {
TurnStatus::Failed {
error: TurnError {
message: "b1".to_string(),
codex_error_code: None,
codex_error_info: None,
}
}
);
codex-rs/app-server/src/codex_message_processor.rs
-56
View File
@@ -92,7 +92,6 @@ use codex_app_server_protocol::TurnStatus;
use codex_app_server_protocol::UserInfoResponse;
use codex_app_server_protocol::UserInput as V2UserInput;
use codex_app_server_protocol::UserSavedConfig;
use codex_app_server_protocol::WindowsWorldWritableWarningNotification;
use codex_app_server_protocol::build_turns_from_event_msgs;
use codex_backend_client::Client as BackendClient;
use codex_core::AuthManager;
@@ -1277,10 +1276,6 @@ impl CodexMessageProcessor {
return;
}
};
if cfg!(windows) && config.features.enabled(Feature::WindowsSandbox) {
self.handle_windows_world_writable_warning(config.cwd.clone())
.await;
}
match self.conversation_manager.new_conversation(config).await {
Ok(conversation_id) => {
@@ -2000,10 +1995,6 @@ impl CodexMessageProcessor {
return;
}
};
if cfg!(windows) && config.features.enabled(Feature::WindowsSandbox) {
self.handle_windows_world_writable_warning(config.cwd.clone())
.await;
}
let conversation_history = if let Some(path) = path {
match RolloutRecorder::get_rollout_history(&path).await {
@@ -2862,53 +2853,6 @@ impl CodexMessageProcessor {
Err(_) => None,
}
}
/// On Windows, when using the experimental sandbox, we need to warn the user about world-writable directories.
async fn handle_windows_world_writable_warning(&self, cwd: PathBuf) {
if !cfg!(windows) {
return;
}
if !self.config.features.enabled(Feature::WindowsSandbox) {
return;
}
if !matches!(
self.config.sandbox_policy,
codex_protocol::protocol::SandboxPolicy::WorkspaceWrite { .. }
| codex_protocol::protocol::SandboxPolicy::ReadOnly
) {
return;
}
if self
.config
.notices
.hide_world_writable_warning
.unwrap_or(false)
{
return;
}
// This function is stubbed out to return None on non-Windows platforms
if let Some((sample_paths, extra_count, failed_scan)) =
codex_windows_sandbox::world_writable_warning_details(
self.config.codex_home.as_path(),
cwd,
)
{
tracing::warn!("world writable warning: {sample_paths:?} {extra_count} {failed_scan}");
self.outgoing
.send_server_notification(ServerNotification::WindowsWorldWritableWarning(
WindowsWorldWritableWarningNotification {
sample_paths,
extra_count,
failed_scan,
},
))
.await;
}
}
}
async fn derive_config_from_params(
codex-rs/app-server/tests/common/Cargo.toml
+2
View File
@@ -24,3 +24,5 @@ tokio = { workspace = true, features = [
] }
uuid = { workspace = true }
wiremock = { workspace = true }
core_test_support = { path = "../../../core/tests/common" }
shlex = { workspace = true }
codex-rs/app-server/tests/common/lib.rs
+3 -1
View File
@@ -9,12 +9,14 @@ pub use auth_fixtures::ChatGptIdTokenClaims;
pub use auth_fixtures::encode_id_token;
pub use auth_fixtures::write_chatgpt_auth;
use codex_app_server_protocol::JSONRPCResponse;
pub use core_test_support::format_with_current_shell;
pub use core_test_support::format_with_current_shell_display;
pub use mcp_process::McpProcess;
pub use mock_model_server::create_mock_chat_completions_server;
pub use mock_model_server::create_mock_chat_completions_server_unchecked;
pub use responses::create_apply_patch_sse_response;
pub use responses::create_final_assistant_message_sse_response;
pub use responses::create_shell_sse_response;
pub use responses::create_shell_command_sse_response;
pub use rollout::create_fake_rollout;
use serde::de::DeserializeOwned;
codex-rs/app-server/tests/common/responses.rs
+10 -9
View File
@@ -1,17 +1,18 @@
use serde_json::json;
use std::path::Path;
pub fn create_shell_sse_response(
pub fn create_shell_command_sse_response(
command: Vec<String>,
workdir: Option<&Path>,
timeout_ms: Option<u64>,
call_id: &str,
) -> anyhow::Result<String> {
// The `arguments`` for the `shell` tool is a serialized JSON object.
// The `arguments` for the `shell_command` tool is a serialized JSON object.
let command_str = shlex::try_join(command.iter().map(String::as_str))?;
let tool_call_arguments = serde_json::to_string(&json!({
"command": command,
"command": command_str,
"workdir": workdir.map(|w| w.to_string_lossy()),
"timeout": timeout_ms
"timeout_ms": timeout_ms
}))?;
let tool_call = json!({
"choices": [
@@ -21,7 +22,7 @@ pub fn create_shell_sse_response(
{
"id": call_id,
"function": {
"name": "shell",
"name": "shell_command",
"arguments": tool_call_arguments
}
}
@@ -62,10 +63,10 @@ pub fn create_apply_patch_sse_response(
patch_content: &str,
call_id: &str,
) -> anyhow::Result<String> {
// Use shell command to call apply_patch with heredoc format
let shell_command = format!("apply_patch <<'EOF'\n{patch_content}\nEOF");
// Use shell_command to call apply_patch with heredoc format
let command = format!("apply_patch <<'EOF'\n{patch_content}\nEOF");
let tool_call_arguments = serde_json::to_string(&json!({
"command": ["bash", "-lc", shell_command]
"command": command
}))?;
let tool_call = json!({
@@ -76,7 +77,7 @@ pub fn create_apply_patch_sse_response(
{
"id": call_id,
"function": {
"name": "shell",
"name": "shell_command",
"arguments": tool_call_arguments
}
}
codex-rs/app-server/tests/suite/codex_message_processor_flow.rs
+12 -27
View File
@@ -2,7 +2,8 @@ use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::create_final_assistant_message_sse_response;
use app_test_support::create_mock_chat_completions_server;
use app_test_support::create_shell_sse_response;
use app_test_support::create_shell_command_sse_response;
use app_test_support::format_with_current_shell;
use app_test_support::to_response;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::AddConversationSubscriptionResponse;
@@ -56,7 +57,7 @@ async fn test_codex_jsonrpc_conversation_flow() -> Result<()> {
// Create a mock model server that immediately ends each turn.
// Two turns are expected: initial session configure + one user message.
let responses = vec![
create_shell_sse_response(
create_shell_command_sse_response(
vec!["ls".to_string()],
Some(&working_directory),
Some(5000),
@@ -175,7 +176,7 @@ async fn test_send_user_turn_changes_approval_policy_behavior() -> Result<()> {
// Mock server will request a python shell call for the first and second turn, then finish.
let responses = vec![
create_shell_sse_response(
create_shell_command_sse_response(
vec![
"python3".to_string(),
"-c".to_string(),
@@ -186,7 +187,7 @@ async fn test_send_user_turn_changes_approval_policy_behavior() -> Result<()> {
"call1",
)?,
create_final_assistant_message_sse_response("done 1")?,
create_shell_sse_response(
create_shell_command_sse_response(
vec![
"python3".to_string(),
"-c".to_string(),
@@ -267,11 +268,7 @@ async fn test_send_user_turn_changes_approval_policy_behavior() -> Result<()> {
ExecCommandApprovalParams {
conversation_id,
call_id: "call1".to_string(),
command: vec![
"python3".to_string(),
"-c".to_string(),
"print(42)".to_string(),
],
command: format_with_current_shell("python3 -c 'print(42)'"),
cwd: working_directory.clone(),
reason: None,
risk: None,
@@ -353,23 +350,15 @@ async fn test_send_user_turn_updates_sandbox_and_cwd_between_turns() -> Result<(
std::fs::create_dir(&second_cwd)?;
let responses = vec![
create_shell_sse_response(
vec![
"bash".to_string(),
"-lc".to_string(),
"echo first turn".to_string(),
],
create_shell_command_sse_response(
vec!["echo".to_string(), "first".to_string(), "turn".to_string()],
None,
Some(5000),
"call-first",
)?,
create_final_assistant_message_sse_response("done first")?,
create_shell_sse_response(
vec![
"bash".to_string(),
"-lc".to_string(),
"echo second turn".to_string(),
],
create_shell_command_sse_response(
vec!["echo".to_string(), "second".to_string(), "turn".to_string()],
None,
Some(5000),
"call-second",
@@ -481,13 +470,9 @@ async fn test_send_user_turn_updates_sandbox_and_cwd_between_turns() -> Result<(
exec_begin.cwd, second_cwd,
"exec turn should run from updated cwd"
);
let expected_command = format_with_current_shell("echo second turn");
assert_eq!(
exec_begin.command,
vec![
"bash".to_string(),
"-lc".to_string(),
"echo second turn".to_string()
],
exec_begin.command, expected_command,
"exec turn should run expected command"
);
codex-rs/app-server/tests/suite/interrupt.rs
+2 -2
View File
@@ -19,7 +19,7 @@ use tokio::time::timeout;
use app_test_support::McpProcess;
use app_test_support::create_mock_chat_completions_server;
use app_test_support::create_shell_sse_response;
use app_test_support::create_shell_command_sse_response;
use app_test_support::to_response;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
@@ -56,7 +56,7 @@ async fn shell_command_interruption() -> anyhow::Result<()> {
std::fs::create_dir(&working_directory)?;
// Create mock server with a single SSE response: the long sleep command
let server = create_mock_chat_completions_server(vec![create_shell_sse_response(
let server = create_mock_chat_completions_server(vec![create_shell_command_sse_response(
shell_command.clone(),
Some(&working_directory),
Some(10_000), // 10 seconds timeout in ms
codex-rs/app-server/tests/suite/v2/turn_interrupt.rs
+2 -2
View File
@@ -3,7 +3,7 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::create_mock_chat_completions_server;
use app_test_support::create_shell_sse_response;
use app_test_support::create_shell_command_sse_response;
use app_test_support::to_response;
use codex_app_server_protocol::JSONRPCNotification;
use codex_app_server_protocol::JSONRPCResponse;
@@ -41,7 +41,7 @@ async fn turn_interrupt_aborts_running_turn() -> Result<()> {
std::fs::create_dir(&working_directory)?;
// Mock server: long-running shell command then (after abort) nothing else needed.
let server = create_mock_chat_completions_server(vec![create_shell_sse_response(
let server = create_mock_chat_completions_server(vec![create_shell_command_sse_response(
shell_command.clone(),
Some(&working_directory),
Some(10_000),
codex-rs/app-server/tests/suite/v2/turn_start.rs
+18 -16
View File
@@ -4,7 +4,8 @@ use app_test_support::create_apply_patch_sse_response;
use app_test_support::create_final_assistant_message_sse_response;
use app_test_support::create_mock_chat_completions_server;
use app_test_support::create_mock_chat_completions_server_unchecked;
use app_test_support::create_shell_sse_response;
use app_test_support::create_shell_command_sse_response;
use app_test_support::format_with_current_shell_display;
use app_test_support::to_response;
use codex_app_server_protocol::ApprovalDecision;
use codex_app_server_protocol::CommandExecutionStatus;
@@ -203,7 +204,7 @@ async fn turn_start_exec_approval_toggle_v2() -> Result<()> {
// Mock server: first turn requests a shell call (elicitation), then completes.
// Second turn same, but we'll set approval_policy=never to avoid elicitation.
let responses = vec![
create_shell_sse_response(
create_shell_command_sse_response(
vec![
"python3".to_string(),
"-c".to_string(),
@@ -214,7 +215,7 @@ async fn turn_start_exec_approval_toggle_v2() -> Result<()> {
"call1",
)?,
create_final_assistant_message_sse_response("done 1")?,
create_shell_sse_response(
create_shell_command_sse_response(
vec![
"python3".to_string(),
"-c".to_string(),
@@ -343,23 +344,15 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
std::fs::create_dir(&second_cwd)?;
let responses = vec![
create_shell_sse_response(
vec![
"bash".to_string(),
"-lc".to_string(),
"echo first turn".to_string(),
],
create_shell_command_sse_response(
vec!["echo".to_string(), "first".to_string(), "turn".to_string()],
None,
Some(5000),
"call-first",
)?,
create_final_assistant_message_sse_response("done first")?,
create_shell_sse_response(
vec![
"bash".to_string(),
"-lc".to_string(),
"echo second turn".to_string(),
],
create_shell_command_sse_response(
vec!["echo".to_string(), "second".to_string(), "turn".to_string()],
None,
Some(5000),
"call-second",
@@ -465,7 +458,8 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
unreachable!("loop ensures we break on command execution items");
};
assert_eq!(cwd, second_cwd);
assert_eq!(command, "bash -lc 'echo second turn'");
let expected_command = format_with_current_shell_display("echo second turn");
assert_eq!(command, expected_command);
assert_eq!(status, CommandExecutionStatus::InProgress);
timeout(
@@ -480,6 +474,10 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
#[tokio::test]
async fn turn_start_file_change_approval_v2() -> Result<()> {
skip_if_no_network!(Ok(()));
if cfg!(windows) {
// TODO apply_patch approvals are not parsed from powershell commands yet
return Ok(());
}
let tmp = TempDir::new()?;
let codex_home = tmp.path().join("codex_home");
@@ -626,6 +624,10 @@ async fn turn_start_file_change_approval_v2() -> Result<()> {
#[tokio::test]
async fn turn_start_file_change_approval_decline_v2() -> Result<()> {
skip_if_no_network!(Ok(()));
if cfg!(windows) {
// TODO apply_patch approvals are not parsed from powershell commands yet
return Ok(());
}
let tmp = TempDir::new()?;
let codex_home = tmp.path().join("codex_home");
codex-rs/apply-patch/src/lib.rs
+15 -7
View File
@@ -30,6 +30,7 @@ pub use standalone_executable::main;
pub const APPLY_PATCH_TOOL_INSTRUCTIONS: &str = include_str!("../apply_patch_tool_instructions.md");
const APPLY_PATCH_COMMANDS: [&str; 2] = ["apply_patch", "applypatch"];
const APPLY_PATCH_SHELLS: [&str; 3] = ["bash", "zsh", "sh"];
#[derive(Debug, Error, PartialEq)]
pub enum ApplyPatchError {
@@ -96,6 +97,13 @@ pub struct ApplyPatchArgs {
pub workdir: Option<String>,
}
fn shell_supports_apply_patch(shell: &str) -> bool {
std::path::Path::new(shell)
.file_name()
.and_then(|name| name.to_str())
.is_some_and(|name| APPLY_PATCH_SHELLS.contains(&name))
}
pub fn maybe_parse_apply_patch(argv: &[String]) -> MaybeApplyPatch {
match argv {
// Direct invocation: apply_patch <patch>
@@ -104,7 +112,7 @@ pub fn maybe_parse_apply_patch(argv: &[String]) -> MaybeApplyPatch {
Err(e) => MaybeApplyPatch::PatchParseError(e),
},
// Bash heredoc form: (optional `cd <path> &&`) apply_patch <<'EOF' ...
[bash, flag, script] if bash == "bash" && flag == "-lc" => {
[shell, flag, script] if shell_supports_apply_patch(shell) && flag == "-lc" => {
match extract_apply_patch_from_bash(script) {
Ok((body, workdir)) => match parse_patch(&body) {
Ok(mut source) => {
@@ -224,12 +232,12 @@ pub fn maybe_parse_apply_patch_verified(argv: &[String], cwd: &Path) -> MaybeApp
);
}
}
[bash, flag, script] if bash == "bash" && flag == "-lc" => {
if parse_patch(script).is_ok() {
return MaybeApplyPatchVerified::CorrectnessError(
ApplyPatchError::ImplicitInvocation,
);
}
[shell, flag, script]
if shell_supports_apply_patch(shell)
&& flag == "-lc"
&& parse_patch(script).is_ok() =>
{
return MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation);
}
_ => {}
}
codex-rs/core/src/codex.rs
+3 -3
View File
@@ -1191,12 +1191,12 @@ impl Session {
message: impl Into<String>,
codex_error: CodexErr,
) {
let codex_error_code = CodexErrorInfo::ResponseStreamDisconnected {
let codex_error_info = CodexErrorInfo::ResponseStreamDisconnected {
http_status_code: codex_error.http_status_code_value(),
};
let event = EventMsg::StreamError(StreamErrorEvent {
message: message.into(),
codex_error_code: Some(codex_error_code),
codex_error_info: Some(codex_error_info),
});
self.send_event(turn_context, event).await;
}
@@ -1689,7 +1689,7 @@ mod handlers {
id: sub_id.clone(),
msg: EventMsg::Error(ErrorEvent {
message: "Failed to shutdown rollout recorder".to_string(),
codex_error_code: Some(CodexErrorInfo::Other),
codex_error_info: Some(CodexErrorInfo::Other),
}),
};
sess.send_event_raw(event).await;
codex-rs/core/src/command_safety/is_safe_command.rs
+14
View File
@@ -267,6 +267,20 @@ mod tests {
}
}
#[test]
fn windows_powershell_full_path_is_safe() {
if !cfg!(windows) {
// Windows only because on Linux path splitting doesn't handle `/` separators properly
return;
}
assert!(is_known_safe_command(&vec_str(&[
r"C:\Program Files\PowerShell\7\pwsh.exe",
"-Command",
"Get-Location",
])));
}
#[test]
fn bash_lc_safe_examples() {
assert!(is_known_safe_command(&vec_str(&["bash", "-lc", "ls"])));
codex-rs/core/src/command_safety/windows_safe_commands.rs
+29 -1
View File
@@ -1,4 +1,5 @@
use shlex::split as shlex_split;
use std::path::Path;
/// On Windows, we conservatively allow only clearly read-only PowerShell invocations
/// that match a small safelist. Anything else (including direct CMD commands) is unsafe.
@@ -131,8 +132,14 @@ fn split_into_commands(tokens: Vec<String>) -> Option<Vec<Vec<String>>> {
/// Returns true when the executable name is one of the supported PowerShell binaries.
fn is_powershell_executable(exe: &str) -> bool {
let executable_name = Path::new(exe)
.file_name()
.and_then(|osstr| osstr.to_str())
.unwrap_or(exe)
.to_ascii_lowercase();
matches!(
exe.to_ascii_lowercase().as_str(),
executable_name.as_str(),
"powershell" | "powershell.exe" | "pwsh" | "pwsh.exe"
)
}
@@ -313,6 +320,27 @@ mod tests {
])));
}
#[test]
fn accepts_full_path_powershell_invocations() {
if !cfg!(windows) {
// Windows only because on Linux path splitting doesn't handle `/` separators properly
return;
}
assert!(is_safe_command_windows(&vec_str(&[
r"C:\Program Files\PowerShell\7\pwsh.exe",
"-NoProfile",
"-Command",
"Get-ChildItem -Path .",
])));
assert!(is_safe_command_windows(&vec_str(&[
r"C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exe",
"-Command",
"Get-Content Cargo.toml",
])));
}
#[test]
fn allows_read_only_pipelines_and_git_usage() {
assert!(is_safe_command_windows(&vec_str(&[
codex-rs/core/src/error.rs
+2 -2
View File
@@ -469,7 +469,7 @@ impl CodexErr {
};
ErrorEvent {
message,
codex_error_code: Some(self.to_codex_protocol_error()),
codex_error_info: Some(self.to_codex_protocol_error()),
}
}
@@ -650,7 +650,7 @@ mod tests {
"prefix: Error while reading the server response: HTTP status client error (429 Too Many Requests) for url (http://example.com/), request id: req-123"
);
assert_eq!(
event.codex_error_code,
event.codex_error_info,
Some(CodexErrorInfo::ResponseStreamConnectionFailed {
http_status_code: Some(429)
})
codex-rs/core/src/event_mapping.rs
+2 -2
View File
@@ -117,7 +117,7 @@ pub fn parse_turn_item(item: &ResponseItem) -> Option<TurnItem> {
..
} => Some(TurnItem::WebSearch(WebSearchItem {
id: id.clone().unwrap_or_default(),
query: query.clone(),
query: query.clone().unwrap_or_default(),
})),
_ => None,
}
@@ -306,7 +306,7 @@ mod tests {
id: Some("ws_1".to_string()),
status: Some("completed".to_string()),
action: WebSearchAction::Search {
query: "weather".to_string(),
query: Some("weather".to_string()),
},
};
codex-rs/core/src/git_info.rs
+14 -4
View File
@@ -825,11 +825,21 @@ mod tests {
.await
.expect("Should collect git info from repo");
let remote_url_output = Command::new("git")
.args(["remote", "get-url", "origin"])
.current_dir(&repo_path)
.output()
.await
.expect("Failed to read remote url");
// Some dev environments rewrite remotes (e.g., force SSH), so compare against
// whatever URL Git reports instead of a fixed placeholder.
let expected_remote = String::from_utf8(remote_url_output.stdout)
.unwrap()
.trim()
.to_string();
// Should have repository URL
assert_eq!(
git_info.repository_url,
Some("https://github.com/example/repo.git".to_string())
);
assert_eq!(git_info.repository_url, Some(expected_remote));
}
#[tokio::test]
codex-rs/core/tests/common/Cargo.toml
+1
View File
@@ -18,3 +18,4 @@ tempfile = { workspace = true }
tokio = { workspace = true, features = ["time"] }
walkdir = { workspace = true }
wiremock = { workspace = true }
shlex = { workspace = true }
codex-rs/core/tests/common/lib.rs
+9
View File
@@ -172,6 +172,15 @@ pub fn sandbox_network_env_var() -> &'static str {
codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR
}
pub fn format_with_current_shell(command: &str) -> Vec<String> {
codex_core::shell::default_user_shell().derive_exec_args(command, true)
}
pub fn format_with_current_shell_display(command: &str) -> String {
let args = format_with_current_shell(command);
shlex::try_join(args.iter().map(String::as_str)).expect("serialize current shell command")
}
pub mod fs_wait {
use anyhow::Result;
use anyhow::anyhow;
codex-rs/core/tests/common/responses.rs
+4 -1
View File
@@ -462,8 +462,11 @@ pub fn ev_apply_patch_function_call(call_id: &str, patch: &str) -> Value {
pub fn ev_shell_command_call(call_id: &str, command: &str) -> Value {
let args = serde_json::json!({ "command": command });
let arguments = serde_json::to_string(&args).expect("serialize shell arguments");
ev_shell_command_call_with_args(call_id, &args)
}
pub fn ev_shell_command_call_with_args(call_id: &str, args: &serde_json::Value) -> Value {
let arguments = serde_json::to_string(args).expect("serialize shell command arguments");
ev_function_call(call_id, "shell_command", &arguments)
}
codex-rs/core/tests/suite/abort_tasks.rs
+16 -15
View File
@@ -17,15 +17,11 @@ use core_test_support::wait_for_event;
use regex_lite::Regex;
use serde_json::json;
/// Integration test: spawn a long‑running shell tool via a mocked Responses SSE
/// Integration test: spawn a long‑running shell_command tool via a mocked Responses SSE
/// function call, then interrupt the session and expect TurnAborted.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn interrupt_long_running_tool_emits_turn_aborted() {
let command = vec![
"bash".to_string(),
"-lc".to_string(),
"sleep 60".to_string(),
];
let command = "sleep 60";
let args = json!({
"command": command,
@@ -33,14 +29,19 @@ async fn interrupt_long_running_tool_emits_turn_aborted() {
})
.to_string();
let body = sse(vec![
ev_function_call("call_sleep", "shell", &args),
ev_function_call("call_sleep", "shell_command", &args),
ev_completed("done"),
]);
let server = start_mock_server().await;
mount_sse_once(&server, body).await;
let codex = test_codex().build(&server).await.unwrap().codex;
let codex = test_codex()
.with_model("gpt-5.1")
.build(&server)
.await
.unwrap()
.codex;
// Kick off a turn that triggers the function call.
codex
@@ -67,11 +68,7 @@ async fn interrupt_long_running_tool_emits_turn_aborted() {
/// responses server, and ensures the model receives the synthesized abort.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn interrupt_tool_records_history_entries() {
let command = vec![
"bash".to_string(),
"-lc".to_string(),
"sleep 60".to_string(),
];
let command = "sleep 60";
let call_id = "call-history";
let args = json!({
@@ -81,7 +78,7 @@ async fn interrupt_tool_records_history_entries() {
.to_string();
let first_body = sse(vec![
ev_response_created("resp-history"),
ev_function_call(call_id, "shell", &args),
ev_function_call(call_id, "shell_command", &args),
ev_completed("resp-history"),
]);
let follow_up_body = sse(vec![
@@ -92,7 +89,11 @@ async fn interrupt_tool_records_history_entries() {
let server = start_mock_server().await;
let response_mock = mount_sse_sequence(&server, vec![first_body, follow_up_body]).await;
let fixture = test_codex().build(&server).await.unwrap();
let fixture = test_codex()
.with_model("gpt-5.1")
.build(&server)
.await
.unwrap();
let codex = Arc::clone(&fixture.codex);
codex
codex-rs/core/tests/suite/apply_patch_cli.rs
+7 -17
View File
@@ -667,7 +667,7 @@ async fn apply_patch_cli_verification_failure_has_no_side_effects(
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_shell_heredoc_with_cd_updates_relative_workdir() -> Result<()> {
async fn apply_patch_shell_command_heredoc_with_cd_updates_relative_workdir() -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness_with(|config| {
@@ -684,14 +684,11 @@ async fn apply_patch_shell_heredoc_with_cd_updates_relative_workdir() -> Result<
let script = "cd sub && apply_patch <<'EOF'\n*** Begin Patch\n*** Update File: in_sub.txt\n@@\n-before\n+after\n*** End Patch\nEOF\n";
let call_id = "shell-heredoc-cd";
let args = json!({
"command": ["bash", "-lc", script],
"timeout_ms": 5_000,
});
let args = json!({ "command": script, "timeout_ms": 5_000 });
let bodies = vec![
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
sse(vec![
@@ -706,14 +703,14 @@ async fn apply_patch_shell_heredoc_with_cd_updates_relative_workdir() -> Result<
let out = harness.function_call_stdout(call_id).await;
assert!(
out.contains("Success."),
"expected successful apply_patch invocation via shell: {out}"
"expected successful apply_patch invocation via shell_command: {out}"
);
assert_eq!(fs::read_to_string(&target)?, "after\n");
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_shell_failure_propagates_error_and_skips_diff() -> Result<()> {
async fn apply_patch_shell_command_failure_propagates_error_and_skips_diff() -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness_with(|config| {
@@ -730,14 +727,11 @@ async fn apply_patch_shell_failure_propagates_error_and_skips_diff() -> Result<(
let script = "apply_patch <<'EOF'\n*** Begin Patch\n*** Update File: invalid.txt\n@@\n-nope\n+changed\n*** End Patch\nEOF\n";
let call_id = "shell-apply-failure";
let args = json!({
"command": ["bash", "-lc", script],
"timeout_ms": 5_000,
});
let args = json!({ "command": script, "timeout_ms": 5_000 });
let bodies = vec![
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
sse(vec![
@@ -780,10 +774,6 @@ async fn apply_patch_shell_failure_propagates_error_and_skips_diff() -> Result<(
);
let out = harness.function_call_stdout(call_id).await;
assert!(
out.contains("apply_patch verification failed"),
"expected verification failure message"
);
assert!(
out.contains("Failed to find expected lines in"),
"expected failure diagnostics: {out}"
codex-rs/core/tests/suite/approvals.rs
+54 -50
View File
@@ -71,7 +71,7 @@ enum ActionKind {
response_body: &'static str,
},
RunCommand {
command: &'static [&'static str],
command: &'static str,
},
RunUnifiedExecCommand {
command: &'static str,
@@ -97,20 +97,12 @@ impl ActionKind {
server: &MockServer,
call_id: &str,
with_escalated_permissions: bool,
) -> Result<(Value, Option<Vec<String>>)> {
) -> Result<(Value, Option<String>)> {
match self {
ActionKind::WriteFile { target, content } => {
let (path, _) = target.resolve_for_patch(test);
let _ = fs::remove_file(&path);
let command = vec![
"/bin/sh".to_string(),
"-c".to_string(),
format!(
"printf {content:?} > {path:?} && cat {path:?}",
content = content,
path = path
),
];
let command = format!("printf {content:?} > {path:?} && cat {path:?}");
let event = shell_event(call_id, &command, 1_000, with_escalated_permissions)?;
Ok((event, Some(command)))
}
@@ -127,21 +119,18 @@ impl ActionKind {
.await;
let url = format!("{}{}", server.uri(), endpoint);
let escaped_url = url.replace('\'', "\\'");
let script = format!(
"import sys\nimport urllib.request\nurl = {url:?}\ntry:\n data = urllib.request.urlopen(url, timeout=2).read().decode()\n print('OK:' + data.strip())\nexcept Exception as exc:\n print('ERR:' + exc.__class__.__name__)\n sys.exit(1)",
"import sys\nimport urllib.request\nurl = '{escaped_url}'\ntry:\n data = urllib.request.urlopen(url, timeout=2).read().decode()\n print('OK:' + data.strip())\nexcept Exception as exc:\n print('ERR:' + exc.__class__.__name__)\n sys.exit(1)",
);
let command = vec!["python3".to_string(), "-c".to_string(), script];
let command = format!("python3 -c \"{script}\"");
let event = shell_event(call_id, &command, 1_000, with_escalated_permissions)?;
Ok((event, Some(command)))
}
ActionKind::RunCommand { command } => {
let command: Vec<String> = command
.iter()
.map(std::string::ToString::to_string)
.collect();
let event = shell_event(call_id, &command, 1_000, with_escalated_permissions)?;
Ok((event, Some(command)))
let event = shell_event(call_id, command, 1_000, with_escalated_permissions)?;
Ok((event, Some(command.to_string())))
}
ActionKind::RunUnifiedExecCommand {
command,
@@ -154,14 +143,7 @@ impl ActionKind {
with_escalated_permissions,
*justification,
)?;
Ok((
event,
Some(vec![
"/bin/bash".to_string(),
"-lc".to_string(),
command.to_string(),
]),
))
Ok((event, Some(command.to_string())))
}
ActionKind::ApplyPatchFunction { target, content } => {
let (path, patch_path) = target.resolve_for_patch(test);
@@ -185,19 +167,19 @@ fn build_add_file_patch(patch_path: &str, content: &str) -> String {
format!("*** Begin Patch\n*** Add File: {patch_path}\n+{content}\n*** End Patch\n")
}
fn shell_apply_patch_command(patch: &str) -> Vec<String> {
fn shell_apply_patch_command(patch: &str) -> String {
let mut script = String::from("apply_patch <<'PATCH'\n");
script.push_str(patch);
if !patch.ends_with('\n') {
script.push('\n');
}
script.push_str("PATCH\n");
vec!["bash".to_string(), "-lc".to_string(), script]
script
}
fn shell_event(
call_id: &str,
command: &[String],
command: &str,
timeout_ms: u64,
with_escalated_permissions: bool,
) -> Result<Value> {
@@ -209,7 +191,7 @@ fn shell_event(
args["with_escalated_permissions"] = json!(true);
}
let args_str = serde_json::to_string(&args)?;
Ok(ev_function_call(call_id, "shell", &args_str))
Ok(ev_function_call(call_id, "shell_command", &args_str))
}
fn exec_command_event(
@@ -296,7 +278,10 @@ impl Expectation {
}
Expectation::FileCreatedNoExitCode { target, content } => {
let (path, _) = target.resolve_for_patch(test);
assert_eq!(result.exit_code, None, "expected no exit code for {path:?}");
assert!(
result.exit_code.is_none() || result.exit_code == Some(0),
"expected no exit code for {path:?}",
);
assert!(
result.stdout.contains(content),
"stdout missing {content:?}: {}",
@@ -385,8 +370,8 @@ impl Expectation {
);
}
Expectation::NetworkSuccessNoExitCode { body_contains } => {
assert_eq!(
result.exit_code, None,
assert!(
result.exit_code.is_none() || result.exit_code == Some(0),
"expected no exit code for successful network call: {}",
result.stdout
);
@@ -433,8 +418,8 @@ impl Expectation {
);
}
Expectation::CommandSuccessNoExitCode { stdout_contains } => {
assert_eq!(
result.exit_code, None,
assert!(
result.exit_code.is_none() || result.exit_code == Some(0),
"expected no exit code for trusted command: {}",
result.stdout
);
@@ -531,10 +516,18 @@ fn parse_result(item: &Value) -> CommandResult {
CommandResult { exit_code, stdout }
}
Err(_) => {
let structured = Regex::new(r"(?s)^Exit code:\s*(-?\d+).*?Output:\n(.*)$").unwrap();
let regex =
Regex::new(r"(?s)^.*?Process exited with code (\d+)\n.*?Output:\n(.*)$").unwrap();
// parse freeform output
if let Some(captures) = regex.captures(output_str) {
if let Some(captures) = structured.captures(output_str) {
let exit_code = captures.get(1).unwrap().as_str().parse::<i64>().unwrap();
let output = captures.get(2).unwrap().as_str();
CommandResult {
exit_code: Some(exit_code),
stdout: output.to_string(),
}
} else if let Some(captures) = regex.captures(output_str) {
let exit_code = captures.get(1).unwrap().as_str().parse::<i64>().unwrap();
let output = captures.get(2).unwrap().as_str();
CommandResult {
@@ -553,7 +546,7 @@ fn parse_result(item: &Value) -> CommandResult {
async fn expect_exec_approval(
test: &TestCodex,
expected_command: &[String],
expected_command: &str,
) -> ExecApprovalRequestEvent {
let event = wait_for_event(&test.codex, |event| {
matches!(
@@ -565,7 +558,12 @@ async fn expect_exec_approval(
match event {
EventMsg::ExecApprovalRequest(approval) => {
assert_eq!(approval.command, expected_command);
let last_arg = approval
.command
.last()
.map(std::string::String::as_str)
.unwrap_or_default();
assert_eq!(last_arg, expected_command);
approval
}
EventMsg::TaskComplete(_) => panic!("expected approval request before completion"),
@@ -660,7 +658,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreatedNoExitCode {
expectation: Expectation::FileCreated {
target: TargetPath::OutsideWorkspace("dfa_on_request_5_1.txt"),
content: "danger-on-request",
},
@@ -702,7 +700,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
approval_policy: UnlessTrusted,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::RunCommand {
command: &["echo", "trusted-unless"],
command: "echo trusted-unless",
},
with_escalated_permissions: false,
features: vec![],
@@ -717,7 +715,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
approval_policy: UnlessTrusted,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::RunCommand {
command: &["echo", "trusted-unless"],
command: "echo trusted-unless",
},
with_escalated_permissions: false,
features: vec![],
@@ -880,7 +878,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::RunCommand {
command: &["echo", "trusted-read-only"],
command: "echo trusted-read-only",
},
with_escalated_permissions: false,
features: vec![],
@@ -895,7 +893,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::RunCommand {
command: &["echo", "trusted-read-only"],
command: "echo trusted-read-only",
},
with_escalated_permissions: false,
features: vec![],
@@ -1020,7 +1018,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
},
ScenarioSpec {
name: "apply_patch_shell_requires_patch_approval",
name: "apply_patch_shell_command_requires_patch_approval",
approval_policy: UnlessTrusted,
sandbox_policy: workspace_write(false),
action: ActionKind::ApplyPatchShell {
@@ -1114,7 +1112,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
},
ScenarioSpec {
name: "apply_patch_shell_outside_requires_patch_approval",
name: "apply_patch_shell_command_outside_requires_patch_approval",
approval_policy: OnRequest,
sandbox_policy: workspace_write(false),
action: ActionKind::ApplyPatchShell {
@@ -1229,7 +1227,10 @@ fn scenarios() -> Vec<ScenarioSpec> {
message_contains: if cfg!(target_os = "linux") {
&["Permission denied"]
} else {
&["Permission denied|Operation not permitted|Read-only file system"]
&[
"Permission denied|Operation not permitted|operation not permitted|\
Read-only file system",
]
},
},
},
@@ -1238,7 +1239,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
approval_policy: Never,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::RunCommand {
command: &["echo", "trusted-never"],
command: "echo trusted-never",
},
with_escalated_permissions: false,
features: vec![],
@@ -1373,7 +1374,10 @@ fn scenarios() -> Vec<ScenarioSpec> {
message_contains: if cfg!(target_os = "linux") {
&["Permission denied"]
} else {
&["Permission denied|Operation not permitted|Read-only file system"]
&[
"Permission denied|Operation not permitted|operation not permitted|\
Read-only file system",
]
},
},
},
@@ -1509,7 +1513,7 @@ async fn run_scenario(scenario: &ScenarioSpec) -> Result<()> {
expected_reason,
} => {
let command = expected_command
.as_ref()
.as_deref()
.expect("exec approval requires shell command");
let approval = expect_exec_approval(&test, command).await;
if let Some(expected_reason) = expected_reason {
codex-rs/core/tests/suite/cli_stream.rs
+13 -2
View File
@@ -499,9 +499,20 @@ async fn integration_git_info_unit_test() {
"Git info should contain repository_url"
);
let repo_url = git_info.repository_url.as_ref().unwrap();
// Some hosts rewrite remotes (e.g., github.com → git@github.com), so assert against
// the actual remote reported by git instead of a static URL.
let expected_remote_url = std::process::Command::new("git")
.args(["remote", "get-url", "origin"])
.current_dir(&git_repo)
.output()
.unwrap();
let expected_remote_url = String::from_utf8(expected_remote_url.stdout)
.unwrap()
.trim()
.to_string();
assert_eq!(
repo_url, "https://github.com/example/integration-test.git",
"Repository URL should match what we configured"
repo_url, &expected_remote_url,
"Repository URL should match git remote get-url output"
);
println!("✅ Git info collection test passed!");
codex-rs/core/tests/suite/client.rs
+1 -1
View File
@@ -992,7 +992,7 @@ async fn azure_responses_request_includes_store_and_reasoning_ids() {
id: Some("web-search-id".into()),
status: Some("completed".into()),
action: WebSearchAction::Search {
query: "weather".into(),
query: Some("weather".into()),
},
});
prompt.input.push(ResponseItem::FunctionCall {
codex-rs/core/tests/suite/codex_delegate.rs
+8 -3
View File
@@ -1,3 +1,4 @@
use codex_core::model_family::find_family_for_model;
use codex_core::protocol::AskForApproval;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
@@ -25,17 +26,17 @@ use pretty_assertions::assert_eq;
async fn codex_delegate_forwards_exec_approval_and_proceeds_on_approval() {
skip_if_no_network!();
// Sub-agent turn 1: emit a shell function_call requiring approval, then complete.
// Sub-agent turn 1: emit a shell_command function_call requiring approval, then complete.
let call_id = "call-exec-1";
let args = serde_json::json!({
"command": ["bash", "-lc", "rm -rf delegated"],
"command": "rm -rf delegated",
"timeout_ms": 1000,
"with_escalated_permissions": true,
})
.to_string();
let sse1 = sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &args),
ev_function_call(call_id, "shell_command", &args),
ev_completed("resp-1"),
]);
@@ -61,6 +62,8 @@ async fn codex_delegate_forwards_exec_approval_and_proceeds_on_approval() {
let mut builder = test_codex().with_config(|config| {
config.approval_policy = AskForApproval::OnRequest;
config.sandbox_policy = SandboxPolicy::ReadOnly;
config.model = "gpt-5.1".to_string();
config.model_family = find_family_for_model("gpt-5.1").expect("gpt-5.1 is a valid model");
});
let test = builder.build(&server).await.expect("build test codex");
@@ -138,6 +141,8 @@ async fn codex_delegate_forwards_patch_approval_and_proceeds_on_decision() {
// Use a restricted sandbox so patch approval is required
config.sandbox_policy = SandboxPolicy::ReadOnly;
config.include_apply_patch_tool = true;
config.model = "gpt-5.1".to_string();
config.model_family = find_family_for_model("gpt-5.1").expect("gpt-5.1 is a valid model");
});
let test = builder.build(&server).await.expect("build test codex");
codex-rs/core/tests/suite/exec_policy.rs
+11 -2
View File
@@ -1,6 +1,7 @@
#![allow(clippy::unwrap_used, clippy::expect_used)]
use anyhow::Result;
use codex_core::model_family::find_family_for_model;
use codex_core::protocol::AskForApproval;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
@@ -21,6 +22,11 @@ use std::fs;
#[tokio::test]
async fn execpolicy_blocks_shell_invocation() -> Result<()> {
// TODO execpolicy doesn't parse powershell commands yet
if cfg!(windows) {
return Ok(());
}
let mut builder = test_codex().with_config(|config| {
let policy_path = config.codex_home.join("policy").join("policy.codexpolicy");
fs::create_dir_all(
@@ -34,13 +40,16 @@ async fn execpolicy_blocks_shell_invocation() -> Result<()> {
r#"prefix_rule(pattern=["echo"], decision="forbidden")"#,
)
.expect("write policy file");
config.model = "gpt-5.1".to_string();
config.model_family =
find_family_for_model("gpt-5.1").expect("gpt-5.1 should have a model family");
});
let server = start_mock_server().await;
let test = builder.build(&server).await?;
let call_id = "shell-forbidden";
let args = json!({
"command": ["echo", "blocked"],
"command": "echo blocked",
"timeout_ms": 1_000,
});
@@ -48,7 +57,7 @@ async fn execpolicy_blocks_shell_invocation() -> Result<()> {
&server,
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
)
codex-rs/core/tests/suite/tool_parallelism.rs
+12 -11
View File
@@ -146,10 +146,11 @@ async fn non_parallel_tools_run_serially() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let test = test_codex().build(&server).await?;
let mut builder = test_codex().with_model("gpt-5.1");
let test = builder.build(&server).await?;
let shell_args = json!({
"command": ["/bin/sh", "-c", "sleep 0.3"],
"command": "sleep 0.3",
"timeout_ms": 1_000,
});
let args_one = serde_json::to_string(&shell_args)?;
@@ -157,8 +158,8 @@ async fn non_parallel_tools_run_serially() -> anyhow::Result<()> {
let first_response = sse(vec![
json!({"type": "response.created", "response": {"id": "resp-1"}}),
ev_function_call("call-1", "shell", &args_one),
ev_function_call("call-2", "shell", &args_two),
ev_function_call("call-1", "shell_command", &args_one),
ev_function_call("call-2", "shell_command", &args_two),
ev_completed("resp-1"),
]);
let second_response = sse(vec![
@@ -167,7 +168,7 @@ async fn non_parallel_tools_run_serially() -> anyhow::Result<()> {
]);
mount_sse_sequence(&server, vec![first_response, second_response]).await;
let duration = run_turn_and_measure(&test, "run shell twice").await?;
let duration = run_turn_and_measure(&test, "run shell_command twice").await?;
assert_serial_duration(duration);
Ok(())
@@ -185,14 +186,14 @@ async fn mixed_tools_fall_back_to_serial() -> anyhow::Result<()> {
})
.to_string();
let shell_args = serde_json::to_string(&json!({
"command": ["/bin/sh", "-c", "sleep 0.3"],
"command": "sleep 0.3",
"timeout_ms": 1_000,
}))?;
let first_response = sse(vec![
json!({"type": "response.created", "response": {"id": "resp-1"}}),
ev_function_call("call-1", "test_sync_tool", &sync_args),
ev_function_call("call-2", "shell", &shell_args),
ev_function_call("call-2", "shell_command", &shell_args),
ev_completed("resp-1"),
]);
let second_response = sse(vec![
@@ -215,7 +216,7 @@ async fn tool_results_grouped() -> anyhow::Result<()> {
let test = build_codex_with_test_tool(&server).await?;
let shell_args = serde_json::to_string(&json!({
"command": ["/bin/sh", "-c", "echo 'shell output'"],
"command": "echo 'shell output'",
"timeout_ms": 1_000,
}))?;
@@ -223,9 +224,9 @@ async fn tool_results_grouped() -> anyhow::Result<()> {
&server,
sse(vec![
json!({"type": "response.created", "response": {"id": "resp-1"}}),
ev_function_call("call-1", "shell", &shell_args),
ev_function_call("call-2", "shell", &shell_args),
ev_function_call("call-3", "shell", &shell_args),
ev_function_call("call-1", "shell_command", &shell_args),
ev_function_call("call-2", "shell_command", &shell_args),
ev_function_call("call-3", "shell_command", &shell_args),
ev_completed("resp-1"),
]),
)
codex-rs/core/tests/suite/truncation.rs
+54 -75
View File
@@ -98,7 +98,7 @@ async fn truncate_function_error_trims_respond_to_model() -> Result<()> {
Ok(())
}
// Verifies that a standard tool call (shell) exceeding the model formatting
// Verifies that a standard tool call (shell_command) exceeding the model formatting
// limits is truncated before being sent back to the model.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn tool_call_output_configured_limit_chars_type() -> Result<()> {
@@ -106,7 +106,7 @@ async fn tool_call_output_configured_limit_chars_type() -> Result<()> {
let server = start_mock_server().await;
// Use a model that exposes the generic shell tool.
// Use a model that exposes the shell_command tool.
let mut builder = test_codex().with_model("gpt-5.1").with_config(|config| {
config.tool_output_token_limit = Some(100_000);
});
@@ -114,28 +114,22 @@ async fn tool_call_output_configured_limit_chars_type() -> Result<()> {
let fixture = builder.build(&server).await?;
let call_id = "shell-too-large";
let args = if cfg!(windows) {
serde_json::json!({
"command": [
"powershell",
"-Command",
"for ($i=1; $i -le 100000; $i++) { Write-Output $i }"
],
"timeout_ms": 5_000,
})
let command = if cfg!(windows) {
"for ($i=1; $i -le 100000; $i++) { Write-Output $i }"
} else {
serde_json::json!({
"command": ["/bin/sh", "-c", "seq 1 100000"],
"timeout_ms": 5_000,
})
"seq 1 100000"
};
let args = serde_json::json!({
"command": command,
"timeout_ms": 5_000,
});
// First response: model tells us to run the tool; second: complete the turn.
mount_sse_once(
&server,
sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
responses::ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
responses::ev_completed("resp-1"),
]),
)
@@ -167,7 +161,10 @@ async fn tool_call_output_configured_limit_chars_type() -> Result<()> {
"expected truncated shell output to be plain text"
);
assert_eq!(output.len(), 400097, "we should be almost 100k tokens");
assert!(
(400000..=401000).contains(&output.len()),
"we should be almost 100k tokens"
);
assert!(
!output.contains("tokens truncated"),
@@ -177,7 +174,7 @@ async fn tool_call_output_configured_limit_chars_type() -> Result<()> {
Ok(())
}
// Verifies that a standard tool call (shell) exceeding the model formatting
// Verifies that a standard tool call (shell_command) exceeding the model formatting
// limits is truncated before being sent back to the model.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn tool_call_output_exceeds_limit_truncated_chars_limit() -> Result<()> {
@@ -185,34 +182,28 @@ async fn tool_call_output_exceeds_limit_truncated_chars_limit() -> Result<()> {
let server = start_mock_server().await;
// Use a model that exposes the generic shell tool.
// Use a model that exposes the shell_command tool.
let mut builder = test_codex().with_model("gpt-5.1");
let fixture = builder.build(&server).await?;
let call_id = "shell-too-large";
let args = if cfg!(windows) {
serde_json::json!({
"command": [
"powershell",
"-Command",
"for ($i=1; $i -le 100000; $i++) { Write-Output $i }"
],
"timeout_ms": 5_000,
})
let command = if cfg!(windows) {
"for ($i=1; $i -le 100000; $i++) { Write-Output $i }"
} else {
serde_json::json!({
"command": ["/bin/sh", "-c", "seq 1 100000"],
"timeout_ms": 5_000,
})
"seq 1 100000"
};
let args = serde_json::json!({
"command": command,
"timeout_ms": 5_000,
});
// First response: model tells us to run the tool; second: complete the turn.
mount_sse_once(
&server,
sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
responses::ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
responses::ev_completed("resp-1"),
]),
)
@@ -250,14 +241,14 @@ async fn tool_call_output_exceeds_limit_truncated_chars_limit() -> Result<()> {
let len = output.len();
assert!(
(9_900..=10_000).contains(&len),
(9_900..=10_100).contains(&len),
"expected ~10k chars after truncation, got {len}"
);
Ok(())
}
// Verifies that a standard tool call (shell) exceeding the model formatting
// Verifies that a standard tool call (shell_command) exceeding the model formatting
// limits is truncated before being sent back to the model.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn tool_call_output_exceeds_limit_truncated_for_model() -> Result<()> {
@@ -265,7 +256,7 @@ async fn tool_call_output_exceeds_limit_truncated_for_model() -> Result<()> {
let server = start_mock_server().await;
// Use a model that exposes the generic shell tool.
// Use a model that exposes the shell_command tool.
let mut builder = test_codex().with_config(|config| {
config.model = "gpt-5.1-codex".to_string();
config.model_family =
@@ -274,28 +265,22 @@ async fn tool_call_output_exceeds_limit_truncated_for_model() -> Result<()> {
let fixture = builder.build(&server).await?;
let call_id = "shell-too-large";
let args = if cfg!(windows) {
serde_json::json!({
"command": [
"powershell",
"-Command",
"for ($i=1; $i -le 100000; $i++) { Write-Output $i }"
],
"timeout_ms": 5_000,
})
let command = if cfg!(windows) {
"for ($i=1; $i -le 100000; $i++) { Write-Output $i }"
} else {
serde_json::json!({
"command": ["/bin/sh", "-c", "seq 1 100000"],
"timeout_ms": 5_000,
})
"seq 1 100000"
};
let args = serde_json::json!({
"command": command,
"timeout_ms": 5_000,
});
// First response: model tells us to run the tool; second: complete the turn.
mount_sse_once(
&server,
sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
responses::ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
responses::ev_completed("resp-1"),
]),
)
@@ -345,7 +330,7 @@ $"#;
Ok(())
}
// Ensures shell tool outputs that exceed the line limit are truncated only once.
// Ensures shell_command outputs that exceed the line limit are truncated only once.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn tool_call_output_truncated_only_once() -> Result<()> {
skip_if_no_network!(Ok(()));
@@ -359,27 +344,21 @@ async fn tool_call_output_truncated_only_once() -> Result<()> {
});
let fixture = builder.build(&server).await?;
let call_id = "shell-single-truncation";
let args = if cfg!(windows) {
serde_json::json!({
"command": [
"powershell",
"-Command",
"for ($i=1; $i -le 10000; $i++) { Write-Output $i }"
],
"timeout_ms": 5_000,
})
let command = if cfg!(windows) {
"for ($i=1; $i -le 10000; $i++) { Write-Output $i }"
} else {
serde_json::json!({
"command": ["/bin/sh", "-c", "seq 1 10000"],
"timeout_ms": 5_000,
})
"seq 1 10000"
};
let args = serde_json::json!({
"command": command,
"timeout_ms": 5_000,
});
mount_sse_once(
&server,
sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
responses::ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
responses::ev_completed("resp-1"),
]),
)
@@ -619,7 +598,7 @@ async fn token_policy_marker_reports_tokens() -> Result<()> {
let call_id = "shell-token-marker";
let args = json!({
"command": ["/bin/sh", "-c", "seq 1 150"],
"command": "seq 1 150",
"timeout_ms": 5_000,
});
@@ -627,7 +606,7 @@ async fn token_policy_marker_reports_tokens() -> Result<()> {
&server,
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
)
@@ -650,7 +629,7 @@ async fn token_policy_marker_reports_tokens() -> Result<()> {
.function_call_output_text(call_id)
.context("shell output present")?;
let pattern = r#"(?s)^\{"output":"Total output lines: 150\\n\\n1\\n2\\n3\\n4\\n5\\n.*?…\d+ tokens truncated…7\\n138\\n139\\n140\\n141\\n142\\n143\\n144\\n145\\n146\\n147\\n148\\n149\\n150\\n","metadata":\{"exit_code":0,"duration_seconds":0\.0\}\}$"#;
let pattern = r"(?s)^Exit code: 0\nWall time: [0-9]+(?:\.[0-9]+)? seconds\nTotal output lines: 150\nOutput:\n1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19.*tokens truncated.*129\n130\n131\n132\n133\n134\n135\n136\n137\n138\n139\n140\n141\n142\n143\n144\n145\n146\n147\n148\n149\n150\n$";
assert_regex_match(pattern, &output);
@@ -672,7 +651,7 @@ async fn byte_policy_marker_reports_bytes() -> Result<()> {
let call_id = "shell-byte-marker";
let args = json!({
"command": ["/bin/sh", "-c", "seq 1 150"],
"command": "seq 1 150",
"timeout_ms": 5_000,
});
@@ -680,7 +659,7 @@ async fn byte_policy_marker_reports_bytes() -> Result<()> {
&server,
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
)
@@ -703,16 +682,16 @@ async fn byte_policy_marker_reports_bytes() -> Result<()> {
.function_call_output_text(call_id)
.context("shell output present")?;
let pattern = r#"(?s)^\{"output":"Total output lines: 150\\n\\n1\\n2\\n3\\n4\\n5.*?…\d+ chars truncated…7\\n138\\n139\\n140\\n141\\n142\\n143\\n144\\n145\\n146\\n147\\n148\\n149\\n150\\n","metadata":\{"exit_code":0,"duration_seconds":0\.0\}\}$"#;
let pattern = r"(?s)^Exit code: 0\nWall time: [0-9]+(?:\.[0-9]+)? seconds\nTotal output lines: 150\nOutput:\n1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19.*chars truncated.*129\n130\n131\n132\n133\n134\n135\n136\n137\n138\n139\n140\n141\n142\n143\n144\n145\n146\n147\n148\n149\n150\n$";
assert_regex_match(pattern, &output);
Ok(())
}
// Shell tool output should remain intact when the config opts into a large token budget.
// shell_command output should remain intact when the config opts into a large token budget.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn shell_tool_output_not_truncated_with_custom_limit() -> Result<()> {
async fn shell_command_output_not_truncated_with_custom_limit() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
@@ -726,7 +705,7 @@ async fn shell_tool_output_not_truncated_with_custom_limit() -> Result<()> {
let call_id = "shell-no-trunc";
let args = json!({
"command": ["/bin/sh", "-c", "seq 1 1000"],
"command": "seq 1 1000",
"timeout_ms": 5_000,
});
let expected_body: String = (1..=1000).map(|i| format!("{i}\n")).collect();
@@ -735,7 +714,7 @@ async fn shell_tool_output_not_truncated_with_custom_limit() -> Result<()> {
&server,
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
)
codex-rs/core/tests/suite/user_shell_cmd.rs
+10 -11
View File
@@ -279,23 +279,19 @@ async fn user_shell_command_is_truncated_only_once() -> anyhow::Result<()> {
config.tool_output_token_limit = Some(100);
config.model = "gpt-5.1-codex".to_string();
config.model_family =
find_family_for_model("gpt-5-codex").expect("gpt-5-codex is a model family");
find_family_for_model("gpt-5.1-codex").expect("gpt-5.1-codex is a model family");
});
let fixture = builder.build(&server).await?;
let call_id = "user-shell-double-truncation";
let args = if cfg!(windows) {
serde_json::json!({
"command": [
"powershell",
"-Command",
"for ($i=1; $i -le 2000; $i++) { Write-Output $i }"
],
"command": "for ($i=1; $i -le 2000; $i++) { Write-Output $i }",
"timeout_ms": 5_000,
})
} else {
serde_json::json!({
"command": ["/bin/sh", "-c", "seq 1 2000"],
"command": "seq 1 2000",
"timeout_ms": 5_000,
})
};
@@ -304,7 +300,7 @@ async fn user_shell_command_is_truncated_only_once() -> anyhow::Result<()> {
&server,
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_function_call(call_id, "shell_command", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
)
@@ -319,19 +315,22 @@ async fn user_shell_command_is_truncated_only_once() -> anyhow::Result<()> {
.await;
fixture
.submit_turn_with_policy("trigger big shell output", SandboxPolicy::DangerFullAccess)
.submit_turn_with_policy(
"trigger big shell_command output",
SandboxPolicy::DangerFullAccess,
)
.await?;
let output = mock2
.single_request()
.function_call_output_text(call_id)
.context("function_call_output present for shell call")?;
.context("function_call_output present for shell_command call")?;
let truncation_headers = output.matches("Total output lines:").count();
assert_eq!(
truncation_headers, 1,
"shell output should carry only one truncation header: {output}"
"shell_command output should carry only one truncation header: {output}"
);
Ok(())
codex-rs/exec/tests/event_processor_with_json_output.rs
+3 -3
View File
@@ -540,7 +540,7 @@ fn error_event_produces_error() {
"e1",
EventMsg::Error(codex_core::protocol::ErrorEvent {
message: "boom".to_string(),
codex_error_code: Some(CodexErrorInfo::Other),
codex_error_info: Some(CodexErrorInfo::Other),
}),
));
assert_eq!(
@@ -580,7 +580,7 @@ fn stream_error_event_produces_error() {
"e1",
EventMsg::StreamError(codex_core::protocol::StreamErrorEvent {
message: "retrying".to_string(),
codex_error_code: Some(CodexErrorInfo::Other),
codex_error_info: Some(CodexErrorInfo::Other),
}),
));
assert_eq!(
@@ -599,7 +599,7 @@ fn error_followed_by_task_complete_produces_turn_failed() {
"e1",
EventMsg::Error(ErrorEvent {
message: "boom".to_string(),
codex_error_code: Some(CodexErrorInfo::Other),
codex_error_info: Some(CodexErrorInfo::Other),
}),
);
assert_eq!(
codex-rs/mcp-server/tests/common/Cargo.toml
+2
View File
@@ -23,3 +23,5 @@ tokio = { workspace = true, features = [
"rt-multi-thread",
] }
wiremock = { workspace = true }
core_test_support = { path = "../../../core/tests/common" }
shlex = { workspace = true }
codex-rs/mcp-server/tests/common/lib.rs
+2 -1
View File
@@ -2,12 +2,13 @@ mod mcp_process;
mod mock_model_server;
mod responses;
pub use core_test_support::format_with_current_shell;
pub use mcp_process::McpProcess;
use mcp_types::JSONRPCResponse;
pub use mock_model_server::create_mock_chat_completions_server;
pub use responses::create_apply_patch_sse_response;
pub use responses::create_final_assistant_message_sse_response;
pub use responses::create_shell_sse_response;
pub use responses::create_shell_command_sse_response;
use serde::de::DeserializeOwned;
pub fn to_response<T: DeserializeOwned>(response: JSONRPCResponse) -> anyhow::Result<T> {
codex-rs/mcp-server/tests/common/responses.rs
+10 -9
View File
@@ -1,17 +1,18 @@
use serde_json::json;
use std::path::Path;
pub fn create_shell_sse_response(
pub fn create_shell_command_sse_response(
command: Vec<String>,
workdir: Option<&Path>,
timeout_ms: Option<u64>,
call_id: &str,
) -> anyhow::Result<String> {
// The `arguments`` for the `shell` tool is a serialized JSON object.
// The `arguments` for the `shell_command` tool is a serialized JSON object.
let command_str = shlex::try_join(command.iter().map(String::as_str))?;
let tool_call_arguments = serde_json::to_string(&json!({
"command": command,
"command": command_str,
"workdir": workdir.map(|w| w.to_string_lossy()),
"timeout": timeout_ms
"timeout_ms": timeout_ms
}))?;
let tool_call = json!({
"choices": [
@@ -21,7 +22,7 @@ pub fn create_shell_sse_response(
{
"id": call_id,
"function": {
"name": "shell",
"name": "shell_command",
"arguments": tool_call_arguments
}
}
@@ -62,10 +63,10 @@ pub fn create_apply_patch_sse_response(
patch_content: &str,
call_id: &str,
) -> anyhow::Result<String> {
// Use shell command to call apply_patch with heredoc format
let shell_command = format!("apply_patch <<'EOF'\n{patch_content}\nEOF");
// Use shell_command to call apply_patch with heredoc format
let command = format!("apply_patch <<'EOF'\n{patch_content}\nEOF");
let tool_call_arguments = serde_json::to_string(&json!({
"command": ["bash", "-lc", shell_command]
"command": command
}))?;
let tool_call = json!({
@@ -76,7 +77,7 @@ pub fn create_apply_patch_sse_response(
{
"id": call_id,
"function": {
"name": "shell",
"name": "shell_command",
"arguments": tool_call_arguments
}
}
codex-rs/mcp-server/tests/suite/codex_tool.rs
+13 -3
View File
@@ -30,7 +30,8 @@ use mcp_test_support::McpProcess;
use mcp_test_support::create_apply_patch_sse_response;
use mcp_test_support::create_final_assistant_message_sse_response;
use mcp_test_support::create_mock_chat_completions_server;
use mcp_test_support::create_shell_sse_response;
use mcp_test_support::create_shell_command_sse_response;
use mcp_test_support::format_with_current_shell;
// Allow ample time on slower CI or under load to avoid flakes.
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(20);
@@ -71,13 +72,16 @@ async fn shell_command_approval_triggers_elicitation() -> anyhow::Result<()> {
"-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 McpHandle {
process: mut mcp_process,
server: _server,
dir: _dir,
} = create_mcp_process(vec![
create_shell_sse_response(
create_shell_command_sse_response(
shell_command.clone(),
Some(workdir_for_shell_function_call.path()),
Some(5_000),
@@ -111,7 +115,7 @@ async fn shell_command_approval_triggers_elicitation() -> anyhow::Result<()> {
)?;
let expected_elicitation_request = create_expected_elicitation_request(
elicitation_request_id.clone(),
shell_command.clone(),
expected_shell_command,
workdir_for_shell_function_call.path(),
codex_request_id.to_string(),
params.codex_event_id.clone(),
@@ -218,6 +222,12 @@ async fn test_patch_approval_triggers_elicitation() {
}
async fn patch_approval_triggers_elicitation() -> anyhow::Result<()> {
if cfg!(windows) {
// powershell apply_patch shell calls are not parsed into apply patch approvals
return Ok(());
}
let cwd = TempDir::new()?;
let test_file = cwd.path().join("destination_file.txt");
std::fs::write(&test_file, "original content\n")?;
codex-rs/protocol/src/models.rs
+83 -1
View File
@@ -230,8 +230,24 @@ pub struct LocalShellExecAction {
#[serde(tag = "type", rename_all = "snake_case")]
pub enum WebSearchAction {
Search {
query: String,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
query: Option<String>,
},
OpenPage {
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
url: Option<String>,
},
FindInPage {
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
url: Option<String>,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pattern: Option<String>,
},
#[serde(other)]
Other,
}
@@ -634,6 +650,72 @@ mod tests {
Ok(())
}
#[test]
fn roundtrips_web_search_call_actions() -> Result<()> {
let cases = vec![
(
r#"{
"type": "web_search_call",
"status": "completed",
"action": {
"type": "search",
"query": "weather seattle"
}
}"#,
WebSearchAction::Search {
query: Some("weather seattle".into()),
},
Some("completed".into()),
),
(
r#"{
"type": "web_search_call",
"status": "open",
"action": {
"type": "open_page",
"url": "https://example.com"
}
}"#,
WebSearchAction::OpenPage {
url: Some("https://example.com".into()),
},
Some("open".into()),
),
(
r#"{
"type": "web_search_call",
"status": "in_progress",
"action": {
"type": "find_in_page",
"url": "https://example.com/docs",
"pattern": "installation"
}
}"#,
WebSearchAction::FindInPage {
url: Some("https://example.com/docs".into()),
pattern: Some("installation".into()),
},
Some("in_progress".into()),
),
];
for (json_literal, expected_action, expected_status) in cases {
let parsed: ResponseItem = serde_json::from_str(json_literal)?;
let expected = ResponseItem::WebSearchCall {
id: None,
status: expected_status.clone(),
action: expected_action.clone(),
};
assert_eq!(parsed, expected);
let serialized = serde_json::to_value(&parsed)?;
let original_value: serde_json::Value = serde_json::from_str(json_literal)?;
assert_eq!(serialized, original_value);
}
Ok(())
}
#[test]
fn deserialize_shell_tool_call_params() -> Result<()> {
let json = r#"{
codex-rs/protocol/src/protocol.rs
+2 -2
View File
@@ -716,7 +716,7 @@ pub struct ExitedReviewModeEvent {
pub struct ErrorEvent {
pub message: String,
#[serde(default)]
pub codex_error_code: Option<CodexErrorInfo>,
pub codex_error_info: Option<CodexErrorInfo>,
}
#[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
@@ -1395,7 +1395,7 @@ pub struct UndoCompletedEvent {
pub struct StreamErrorEvent {
pub message: String,
#[serde(default)]
pub codex_error_code: Option<CodexErrorInfo>,
pub codex_error_info: Option<CodexErrorInfo>,
}
#[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
codex-rs/tui/src/bottom_pane/list_selection_view.rs
+6 -1
View File
@@ -52,6 +52,7 @@ pub(crate) struct SelectionViewParams {
pub is_searchable: bool,
pub search_placeholder: Option<String>,
pub header: Box<dyn Renderable>,
pub initial_selected_idx: Option<usize>,
}
impl Default for SelectionViewParams {
@@ -64,6 +65,7 @@ impl Default for SelectionViewParams {
is_searchable: false,
search_placeholder: None,
header: Box::new(()),
initial_selected_idx: None,
}
}
}
@@ -80,6 +82,7 @@ pub(crate) struct ListSelectionView {
filtered_indices: Vec<usize>,
last_selected_actual_idx: Option<usize>,
header: Box<dyn Renderable>,
initial_selected_idx: Option<usize>,
}
impl ListSelectionView {
@@ -110,6 +113,7 @@ impl ListSelectionView {
filtered_indices: Vec::new(),
last_selected_actual_idx: None,
header,
initial_selected_idx: params.initial_selected_idx,
};
s.apply_filter();
s
@@ -132,7 +136,8 @@ impl ListSelectionView {
(!self.is_searchable)
.then(|| self.items.iter().position(|item| item.is_current))
.flatten()
});
})
.or_else(|| self.initial_selected_idx.take());
if self.is_searchable && !self.search_query.is_empty() {
let query_lower = self.search_query.to_lowercase();
codex-rs/tui/src/chatwidget.rs
+9
View File
@@ -2120,6 +2120,14 @@ impl ChatWidget {
} else {
default_choice
};
let selection_choice = highlight_choice.or(default_choice);
let initial_selected_idx = choices
.iter()
.position(|choice| choice.stored == selection_choice)
.or_else(|| {
selection_choice
.and_then(|effort| choices.iter().position(|choice| choice.display == effort))
});
let mut items: Vec<SelectionItem> = Vec::new();
for choice in choices.iter() {
let effort = choice.display;
@@ -2196,6 +2204,7 @@ impl ChatWidget {
header: Box::new(header),
footer_hint: Some(standard_popup_hint_line()),
items,
initial_selected_idx,
..Default::default()
});
}
codex-rs/tui/src/chatwidget/tests.rs
+1 -1
View File
@@ -2648,7 +2648,7 @@ fn stream_error_updates_status_indicator() {
id: "sub-1".into(),
msg: EventMsg::StreamError(StreamErrorEvent {
message: msg.to_string(),
codex_error_code: Some(CodexErrorInfo::Other),
codex_error_info: Some(CodexErrorInfo::Other),
}),
});
codex-rs/windows-sandbox-rs/src/allow.rs
+36 -1
View File
@@ -18,6 +18,13 @@ pub fn compute_allow_paths(
allow.push(p);
}
};
let include_tmp_env_vars = matches!(
policy,
SandboxPolicy::WorkspaceWrite {
exclude_tmpdir_env_var: false,
..
}
);
if matches!(policy, SandboxPolicy::WorkspaceWrite { .. }) {
add_path(command_cwd.to_path_buf());
@@ -33,7 +40,7 @@ pub fn compute_allow_paths(
}
}
}
if !matches!(policy, SandboxPolicy::ReadOnly) {
if include_tmp_env_vars {
for key in ["TEMP", "TMP"] {
if let Some(v) = env_map.get(key) {
let abs = PathBuf::from(v);
@@ -80,4 +87,32 @@ mod tests {
"additional writable root should be allowed"
);
}
#[test]
fn excludes_tmp_env_vars_when_requested() {
let command_cwd = PathBuf::from(r"C:\Workspace");
let temp_dir = PathBuf::from(r"C:\TempDir");
let _ = fs::create_dir_all(&command_cwd);
let _ = fs::create_dir_all(&temp_dir);
let policy = SandboxPolicy::WorkspaceWrite {
writable_roots: vec![],
network_access: false,
exclude_tmpdir_env_var: true,
exclude_slash_tmp: false,
};
let mut env_map = HashMap::new();
env_map.insert("TEMP".into(), temp_dir.to_string_lossy().to_string());
let allow = compute_allow_paths(&policy, &command_cwd, &command_cwd, &env_map);
assert!(
allow.iter().any(|p| p == &command_cwd),
"command cwd should be allowed"
);
assert!(
!allow.iter().any(|p| p == &temp_dir),
"TEMP should be excluded when exclude_tmpdir_env_var is true"
);
}
}
codex-rs/windows-sandbox-rs/src/lib.rs
+38 -1
View File
@@ -71,6 +71,10 @@ mod windows_impl {
type PipeHandles = ((HANDLE, HANDLE), (HANDLE, HANDLE), (HANDLE, HANDLE));
fn should_apply_network_block(policy: &SandboxPolicy) -> bool {
!policy.has_full_network_access()
}
fn ensure_dir(p: &Path) -> Result<()> {
if let Some(d) = p.parent() {
std::fs::create_dir_all(d)?;
@@ -214,9 +218,12 @@ mod windows_impl {
timeout_ms: Option<u64>,
) -> Result<CaptureResult> {
let policy = parse_policy(policy_json_or_preset)?;
let apply_network_block = should_apply_network_block(&policy);
normalize_null_device_env(&mut env_map);
ensure_non_interactive_pager(&mut env_map);
apply_no_network_to_env(&mut env_map)?;
if apply_network_block {
apply_no_network_to_env(&mut env_map)?;
}
ensure_codex_home_exists(codex_home)?;
let current_dir = cwd.to_path_buf();
@@ -447,6 +454,36 @@ mod windows_impl {
timed_out,
})
}
#[cfg(test)]
mod tests {
use super::should_apply_network_block;
use crate::policy::SandboxPolicy;
fn workspace_policy(network_access: bool) -> SandboxPolicy {
SandboxPolicy::WorkspaceWrite {
writable_roots: Vec::new(),
network_access,
exclude_tmpdir_env_var: false,
exclude_slash_tmp: false,
}
}
#[test]
fn applies_network_block_when_access_is_disabled() {
assert!(should_apply_network_block(&workspace_policy(false)));
}
#[test]
fn skips_network_block_when_access_is_allowed() {
assert!(!should_apply_network_block(&workspace_policy(true)));
}
#[test]
fn applies_network_block_for_read_only() {
assert!(should_apply_network_block(&SandboxPolicy::ReadOnly));
}
}
}
#[cfg(not(target_os = "windows"))]