| 1 | //! Cargo test runner tool: `run_tests`. |
| 2 | //! |
| 3 | //! This tool intentionally auto-approves test execution to encourage |
| 4 | //! frequent verification loops while still scoping execution to the workspace. |
| 5 | |
| 6 | use std::path::Path; |
| 7 | use std::process::Command; |
| 8 | |
| 9 | use async_trait::async_trait; |
| 10 | use serde::{Deserialize, Serialize}; |
| 11 | use serde_json::{Value, json}; |
| 12 | |
| 13 | use super::spec::{ |
| 14 | ApprovalRequirement, ToolCapability, ToolContext, ToolError, ToolResult, ToolSpec, |
| 15 | optional_bool, optional_str, |
| 16 | }; |
| 17 | |
| 18 | const MAX_OUTPUT_CHARS: usize = 40_000; |
| 19 | |
| 20 | /// Tool for running `cargo test` in the workspace root. |
| 21 | pub struct RunTestsTool; |
| 22 | |
| 23 | #[derive(Debug, Clone, Serialize, Deserialize)] |
| 24 | struct RunTestsOutput { |
| 25 | success: bool, |
| 26 | exit_code: i32, |
| 27 | stdout: String, |
| 28 | stderr: String, |
| 29 | command: String, |
| 30 | } |
| 31 | |
| 32 | #[async_trait] |
| 33 | impl ToolSpec for RunTestsTool { |
| 34 | fn name(&self) -> &'static str { |
| 35 | "run_tests" |
| 36 | } |
| 37 | |
| 38 | fn description(&self) -> &'static str { |
| 39 | "Run `cargo test` in the workspace root with optional extra arguments." |
| 40 | } |
| 41 | |
| 42 | fn input_schema(&self) -> Value { |
| 43 | json!({ |
| 44 | "type": "object", |
| 45 | "properties": { |
| 46 | "args": { |
| 47 | "type": "string", |
| 48 | "description": "Optional extra arguments to pass to `cargo test` (shell-style)." |
| 49 | }, |
| 50 | "all_features": { |
| 51 | "type": "boolean", |
| 52 | "description": "When true, include `--all-features`." |
| 53 | } |
| 54 | }, |
| 55 | "additionalProperties": false |
| 56 | }) |
| 57 | } |
| 58 | |
| 59 | fn capabilities(&self) -> Vec<ToolCapability> { |
| 60 | vec![ToolCapability::ExecutesCode, ToolCapability::Sandboxable] |
| 61 | } |
| 62 | |
| 63 | fn approval_requirement(&self) -> ApprovalRequirement { |
| 64 | // Tests are encouraged, so avoid gating them behind approval. |
| 65 | ApprovalRequirement::Auto |
| 66 | } |
| 67 | |
| 68 | async fn execute(&self, input: Value, context: &ToolContext) -> Result<ToolResult, ToolError> { |
| 69 | let all_features = optional_bool(&input, "all_features", false); |
| 70 | let extra_args = optional_str(&input, "args") |
| 71 | .map(str::trim) |
| 72 | .filter(|s| !s.is_empty()); |
| 73 | |
| 74 | let mut args = vec!["test".to_string()]; |
| 75 | if all_features { |
| 76 | args.push("--all-features".to_string()); |
| 77 | } |
| 78 | if let Some(extra) = extra_args { |
| 79 | let split = shlex::split(extra).ok_or_else(|| { |
| 80 | ToolError::invalid_input("Failed to parse 'args' as shell-style tokens") |
| 81 | })?; |
| 82 | args.extend(split); |
| 83 | } |
| 84 | |
| 85 | let command_str = format_command(&context.workspace, &args); |
| 86 | let output = run_cargo(&context.workspace, &args)?; |
| 87 | |
| 88 | let exit_code = output.status.code().unwrap_or(-1); |
| 89 | let stdout_raw = String::from_utf8_lossy(&output.stdout); |
| 90 | let stderr_raw = String::from_utf8_lossy(&output.stderr); |
| 91 | let stdout = truncate_with_note(&stdout_raw, MAX_OUTPUT_CHARS); |
| 92 | let stderr = truncate_with_note(&stderr_raw, MAX_OUTPUT_CHARS); |
| 93 | |
| 94 | let result = RunTestsOutput { |
| 95 | success: output.status.success(), |
| 96 | exit_code, |
| 97 | stdout, |
| 98 | stderr, |
| 99 | command: command_str, |
| 100 | }; |
| 101 | |
| 102 | ToolResult::json(&result).map_err(|e| ToolError::execution_failed(e.to_string())) |
| 103 | } |
| 104 | } |
| 105 | |
| 106 | // === Helpers === |
| 107 | |
| 108 | fn run_cargo(workspace: &Path, args: &[String]) -> Result<std::process::Output, ToolError> { |
| 109 | let mut cmd = Command::new("cargo"); |
| 110 | cmd.args(args).current_dir(workspace); |
| 111 | cmd.output().map_err(|e| { |
| 112 | if e.kind() == std::io::ErrorKind::NotFound { |
| 113 | ToolError::not_available("cargo is not installed or not in PATH") |
| 114 | } else { |
| 115 | ToolError::execution_failed(format!("Failed to run cargo: {e}")) |
| 116 | } |
| 117 | }) |
| 118 | } |
| 119 | |
| 120 | fn format_command(workspace: &Path, args: &[String]) -> String { |
| 121 | format!( |
| 122 | "(cd {} && cargo {})", |
| 123 | workspace.display(), |
| 124 | args.iter() |
| 125 | .map(String::as_str) |
| 126 | .collect::<Vec<_>>() |
| 127 | .join(" ") |
| 128 | ) |
| 129 | } |
| 130 | |
| 131 | fn truncate_with_note(text: &str, max_chars: usize) -> String { |
| 132 | if text.chars().count() <= max_chars { |
| 133 | return text.to_string(); |
| 134 | } |
| 135 | let end = char_boundary_index(text, max_chars); |
| 136 | let truncated = &text[..end]; |
| 137 | let omitted_chars = text |
| 138 | .chars() |
| 139 | .count() |
| 140 | .saturating_sub(truncated.chars().count()); |
| 141 | let note = format!( |
| 142 | "\n\n[output truncated to {max_chars} characters; {omitted_chars} characters omitted]" |
| 143 | ); |
| 144 | format!("{truncated}{note}") |
| 145 | } |
| 146 | |
| 147 | fn char_boundary_index(text: &str, max_chars: usize) -> usize { |
| 148 | if max_chars == 0 { |
| 149 | return 0; |
| 150 | } |
| 151 | for (count, (idx, _)) in text.char_indices().enumerate() { |
| 152 | if count == max_chars { |
| 153 | return idx; |
| 154 | } |
| 155 | } |
| 156 | text.len() |
| 157 | } |
| 158 | |
| 159 | #[cfg(test)] |
| 160 | mod tests { |
| 161 | use super::*; |
| 162 | use std::fs; |
| 163 | use std::process::Command; |
| 164 | use tempfile::tempdir; |
| 165 | |
| 166 | fn cargo_available() -> bool { |
| 167 | Command::new("cargo") |
| 168 | .arg("--version") |
| 169 | .output() |
| 170 | .map(|o| o.status.success()) |
| 171 | .unwrap_or(false) |
| 172 | } |
| 173 | |
| 174 | fn init_cargo_project(root: &Path) -> std::path::PathBuf { |
| 175 | let project_dir = root.join("project"); |
| 176 | fs::create_dir_all(&project_dir).expect("create project dir"); |
| 177 | let status = Command::new("cargo") |
| 178 | .args([ |
| 179 | "init", |
| 180 | "--lib", |
| 181 | "--vcs", |
| 182 | "none", |
| 183 | "-q", |
| 184 | "--name", |
| 185 | "eval_project", |
| 186 | ]) |
| 187 | .current_dir(&project_dir) |
| 188 | .status() |
| 189 | .expect("cargo should spawn"); |
| 190 | assert!(status.success(), "cargo init failed"); |
| 191 | project_dir |
| 192 | } |
| 193 | |
| 194 | #[tokio::test] |
| 195 | async fn run_tests_succeeds_on_fresh_project() { |
| 196 | if !cargo_available() { |
| 197 | return; |
| 198 | } |
| 199 | let tmp = tempdir().expect("tempdir"); |
| 200 | let project_dir = init_cargo_project(tmp.path()); |
| 201 | |
| 202 | let ctx = ToolContext::new(&project_dir); |
| 203 | let tool = RunTestsTool; |
| 204 | let result = tool.execute(json!({}), &ctx).await.expect("execute"); |
| 205 | assert!(result.success); |
| 206 | |
| 207 | let parsed: RunTestsOutput = |
| 208 | serde_json::from_str(&result.content).expect("tool result should be json"); |
| 209 | assert!(parsed.success); |
| 210 | assert_eq!(parsed.exit_code, 0); |
| 211 | assert!(parsed.command.contains("cargo test")); |
| 212 | } |
| 213 | |
| 214 | #[tokio::test] |
| 215 | async fn run_tests_reports_failures_without_hard_error() { |
| 216 | if !cargo_available() { |
| 217 | return; |
| 218 | } |
| 219 | let tmp = tempdir().expect("tempdir"); |
| 220 | let project_dir = init_cargo_project(tmp.path()); |
| 221 | |
| 222 | let lib_rs = project_dir.join("src/lib.rs"); |
| 223 | let failing = r#" |
| 224 | pub fn add(a: i32, b: i32) -> i32 { a + b } |
| 225 | |
| 226 | #[cfg(test)] |
| 227 | mod tests { |
| 228 | #[test] |
| 229 | fn fails() { |
| 230 | assert_eq!(2 + 2, 5); |
| 231 | } |
| 232 | } |
| 233 | "#; |
| 234 | fs::write(&lib_rs, failing).expect("write failing test"); |
| 235 | |
| 236 | let ctx = ToolContext::new(&project_dir); |
| 237 | let tool = RunTestsTool; |
| 238 | let result = tool.execute(json!({}), &ctx).await.expect("execute"); |
| 239 | assert!(result.success); |
| 240 | |
| 241 | let parsed: RunTestsOutput = |
| 242 | serde_json::from_str(&result.content).expect("tool result should be json"); |
| 243 | assert!(!parsed.success); |
| 244 | assert_ne!(parsed.exit_code, 0); |
| 245 | } |
| 246 | |
| 247 | #[test] |
| 248 | fn truncation_adds_note() { |
| 249 | let long = "x".repeat(MAX_OUTPUT_CHARS + 128); |
| 250 | let truncated = truncate_with_note(&long, MAX_OUTPUT_CHARS); |
| 251 | assert!(truncated.contains("output truncated")); |
| 252 | } |
| 253 | } |
| 254 |