g3 console initial cut + error doesnt kill auto
This commit is contained in:
270
crates/g3-console/src/process/controller.rs
Normal file
270
crates/g3-console/src/process/controller.rs
Normal file
@@ -0,0 +1,270 @@
|
||||
use anyhow::{anyhow, Context, Result};
|
||||
use std::process::{Command, Stdio};
|
||||
use std::os::unix::process::CommandExt;
|
||||
use std::collections::HashMap;
|
||||
use std::sync::Mutex;
|
||||
use std::path::PathBuf;
|
||||
use sysinfo::{Pid, Signal, System, Process};
|
||||
use tracing::{debug, error, info};
|
||||
use crate::models::LaunchParams;
|
||||
|
||||
pub struct ProcessController {
|
||||
system: System,
|
||||
launch_params: Mutex<HashMap<u32, LaunchParams>>,
|
||||
}
|
||||
|
||||
impl ProcessController {
|
||||
pub fn new() -> Self {
|
||||
Self {
|
||||
system: System::new_all(),
|
||||
launch_params: Mutex::new(HashMap::new()),
|
||||
}
|
||||
}
|
||||
|
||||
pub fn kill_process(&mut self, pid: u32) -> Result<()> {
|
||||
let sysinfo_pid = Pid::from_u32(pid);
|
||||
self.system.refresh_processes();
|
||||
|
||||
if let Some(process) = self.system.process(sysinfo_pid) {
|
||||
info!("Killing process {} ({})", pid, process.name());
|
||||
|
||||
// Try SIGTERM first
|
||||
if process.kill_with(Signal::Term).is_some() {
|
||||
debug!("Sent SIGTERM to process {}", pid);
|
||||
|
||||
// Wait a bit and check if it's still running
|
||||
std::thread::sleep(std::time::Duration::from_secs(2));
|
||||
self.system.refresh_processes();
|
||||
|
||||
if self.system.process(sysinfo_pid).is_some() {
|
||||
// Still running, send SIGKILL
|
||||
if let Some(proc) = self.system.process(sysinfo_pid) {
|
||||
proc.kill_with(Signal::Kill);
|
||||
debug!("Sent SIGKILL to process {}", pid);
|
||||
}
|
||||
}
|
||||
|
||||
Ok(())
|
||||
} else {
|
||||
Err(anyhow!("Failed to send signal to process {}", pid))
|
||||
}
|
||||
} else {
|
||||
Err(anyhow!("Process {} not found", pid))
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(unix)]
|
||||
pub fn launch_g3(
|
||||
&mut self,
|
||||
workspace: &str,
|
||||
provider: &str,
|
||||
model: &str,
|
||||
prompt: &str,
|
||||
autonomous: bool,
|
||||
g3_binary_path: Option<&str>,
|
||||
) -> Result<u32> {
|
||||
let binary = g3_binary_path.unwrap_or("g3");
|
||||
|
||||
let mut cmd = Command::new(binary);
|
||||
cmd.arg("--workspace")
|
||||
.arg(workspace)
|
||||
.arg("--provider")
|
||||
.arg(provider)
|
||||
.arg("--model")
|
||||
.arg(model);
|
||||
|
||||
if autonomous {
|
||||
cmd.arg("--autonomous");
|
||||
}
|
||||
|
||||
cmd.arg(prompt);
|
||||
|
||||
// Run in background with proper detachment
|
||||
cmd.stdout(Stdio::null())
|
||||
.stderr(Stdio::null())
|
||||
.stdin(Stdio::null());
|
||||
|
||||
// Double-fork technique to prevent zombie processes:
|
||||
// 1. Fork once to create intermediate process
|
||||
// 2. Intermediate process forks again and exits immediately
|
||||
// 3. Grandchild is adopted by init (PID 1) which will reap it
|
||||
unsafe {
|
||||
cmd.pre_exec(|| {
|
||||
// Fork again inside the child
|
||||
match libc::fork() {
|
||||
-1 => return Err(std::io::Error::last_os_error()),
|
||||
0 => {
|
||||
// Grandchild: create new session and continue
|
||||
libc::setsid();
|
||||
// Continue execution (this becomes the actual g3 process)
|
||||
}
|
||||
_ => {
|
||||
// Child: exit immediately so parent can reap it
|
||||
libc::_exit(0);
|
||||
}
|
||||
}
|
||||
Ok(())
|
||||
});
|
||||
}
|
||||
|
||||
info!("Launching g3: {:?}", cmd);
|
||||
|
||||
// Spawn and wait for the intermediate process to exit
|
||||
let mut child = cmd.spawn().context("Failed to spawn g3 process")?;
|
||||
let intermediate_pid = child.id();
|
||||
|
||||
// Wait for intermediate process (it will exit immediately after forking)
|
||||
child.wait().context("Failed to wait for intermediate process")?;
|
||||
|
||||
// The actual g3 process is now running as orphan
|
||||
// We need to scan for it by matching workspace and recent start time
|
||||
info!("Scanning for newly launched g3 process in workspace: {}", workspace);
|
||||
|
||||
// Wait a moment for the process to fully start
|
||||
std::thread::sleep(std::time::Duration::from_millis(500));
|
||||
|
||||
// Refresh and scan for the process
|
||||
self.system.refresh_processes();
|
||||
let workspace_path = PathBuf::from(workspace);
|
||||
let mut found_pid = None;
|
||||
|
||||
for (pid, process) in self.system.processes() {
|
||||
let cmd = process.cmd();
|
||||
let cmd_str = cmd.join(" ");
|
||||
|
||||
// Check if this is a g3 process
|
||||
let is_g3 = process.name().contains("g3") || cmd_str.contains("g3");
|
||||
if !is_g3 {
|
||||
continue;
|
||||
}
|
||||
|
||||
// Check if it has our workspace
|
||||
let has_workspace = cmd.iter().any(|arg| {
|
||||
if let Ok(path) = PathBuf::from(arg).canonicalize() {
|
||||
if let Ok(ws) = workspace_path.canonicalize() {
|
||||
return path == ws;
|
||||
}
|
||||
}
|
||||
false
|
||||
});
|
||||
|
||||
if has_workspace {
|
||||
// Check if it's recent (started within last 5 seconds)
|
||||
let now = std::time::SystemTime::now();
|
||||
let start_time = std::time::UNIX_EPOCH + std::time::Duration::from_secs(process.start_time());
|
||||
if let Ok(duration) = now.duration_since(start_time) {
|
||||
if duration.as_secs() < 5 {
|
||||
found_pid = Some(pid.as_u32());
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
let pid = found_pid.unwrap_or(intermediate_pid);
|
||||
|
||||
info!("Launched g3 process with PID {}", pid);
|
||||
|
||||
// Store launch params for restart
|
||||
let params = LaunchParams {
|
||||
workspace: workspace.into(),
|
||||
provider: provider.to_string(),
|
||||
model: model.to_string(),
|
||||
prompt: prompt.to_string(),
|
||||
autonomous,
|
||||
g3_binary_path: g3_binary_path.map(|s| s.to_string()),
|
||||
};
|
||||
|
||||
if let Ok(mut map) = self.launch_params.lock() {
|
||||
map.insert(pid, params);
|
||||
}
|
||||
|
||||
Ok(pid)
|
||||
}
|
||||
|
||||
pub fn get_launch_params(&mut self, pid: u32) -> Option<LaunchParams> {
|
||||
// First check if we have stored params (for console-launched instances)
|
||||
if let Ok(map) = self.launch_params.lock() {
|
||||
if let Some(params) = map.get(&pid) {
|
||||
return Some(params.clone());
|
||||
}
|
||||
}
|
||||
|
||||
// If not found, try to parse from process command line (for detected instances)
|
||||
self.system.refresh_processes();
|
||||
let sysinfo_pid = Pid::from_u32(pid);
|
||||
|
||||
if let Some(process) = self.system.process(sysinfo_pid) {
|
||||
let cmd = process.cmd();
|
||||
return self.parse_launch_params_from_cmd(cmd);
|
||||
}
|
||||
|
||||
None
|
||||
}
|
||||
|
||||
fn parse_launch_params_from_cmd(&self, cmd: &[String]) -> Option<LaunchParams> {
|
||||
let mut workspace = None;
|
||||
let mut provider = None;
|
||||
let mut model = None;
|
||||
let mut prompt = None;
|
||||
let mut autonomous = false;
|
||||
let mut g3_binary_path = None;
|
||||
|
||||
let mut i = 0;
|
||||
while i < cmd.len() {
|
||||
match cmd[i].as_str() {
|
||||
"--workspace" | "-w" if i + 1 < cmd.len() => {
|
||||
workspace = Some(PathBuf::from(&cmd[i + 1]));
|
||||
i += 2;
|
||||
}
|
||||
"--provider" if i + 1 < cmd.len() => {
|
||||
provider = Some(cmd[i + 1].clone());
|
||||
i += 2;
|
||||
}
|
||||
"--model" if i + 1 < cmd.len() => {
|
||||
model = Some(cmd[i + 1].clone());
|
||||
i += 2;
|
||||
}
|
||||
"--autonomous" => {
|
||||
autonomous = true;
|
||||
i += 1;
|
||||
}
|
||||
_ => {
|
||||
// Last non-flag argument is likely the prompt
|
||||
if !cmd[i].starts_with('-') && i == cmd.len() - 1 {
|
||||
prompt = Some(cmd[i].clone());
|
||||
}
|
||||
i += 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Try to determine binary path from cmd[0]
|
||||
if !cmd.is_empty() {
|
||||
let first = &cmd[0];
|
||||
if first.contains("g3") && !first.contains("cargo") {
|
||||
g3_binary_path = Some(first.clone());
|
||||
}
|
||||
}
|
||||
|
||||
// Only return params if we have the minimum required fields
|
||||
if let (Some(ws), Some(prov), Some(mdl), Some(prmt)) = (workspace, provider, model, prompt) {
|
||||
Some(LaunchParams {
|
||||
workspace: ws,
|
||||
provider: prov,
|
||||
model: mdl,
|
||||
prompt: prmt,
|
||||
autonomous,
|
||||
g3_binary_path,
|
||||
})
|
||||
} else {
|
||||
None
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl Default for ProcessController {
|
||||
fn default() -> Self {
|
||||
Self::new()
|
||||
}
|
||||
}
|
||||
172
crates/g3-console/src/process/detector.rs
Normal file
172
crates/g3-console/src/process/detector.rs
Normal file
@@ -0,0 +1,172 @@
|
||||
use crate::models::{ExecutionMethod, Instance, InstanceStatus, InstanceType};
|
||||
use anyhow::{Context, Result};
|
||||
use chrono::{DateTime, Utc};
|
||||
use std::path::PathBuf;
|
||||
use std::process::Command;
|
||||
use sysinfo::{System, Process, Pid};
|
||||
use tracing::{debug, warn};
|
||||
|
||||
pub struct ProcessDetector {
|
||||
system: System,
|
||||
}
|
||||
|
||||
impl ProcessDetector {
|
||||
pub fn new() -> Self {
|
||||
Self {
|
||||
system: System::new_all(),
|
||||
}
|
||||
}
|
||||
|
||||
pub fn detect_instances(&mut self) -> Result<Vec<Instance>> {
|
||||
self.system.refresh_processes();
|
||||
let mut instances = Vec::new();
|
||||
|
||||
// Find all g3 processes
|
||||
for (pid, process) in self.system.processes() {
|
||||
let cmd = process.cmd();
|
||||
if cmd.is_empty() {
|
||||
continue;
|
||||
}
|
||||
|
||||
// Check if this is a g3 process (binary or cargo run)
|
||||
if let Some(instance) = self.parse_g3_process(*pid, process, cmd) {
|
||||
instances.push(instance);
|
||||
}
|
||||
}
|
||||
|
||||
debug!("Detected {} g3 instances", instances.len());
|
||||
Ok(instances)
|
||||
}
|
||||
|
||||
fn parse_g3_process(
|
||||
&self,
|
||||
pid: Pid,
|
||||
process: &Process,
|
||||
cmd: &[String],
|
||||
) -> Option<Instance> {
|
||||
let cmd_str = cmd.join(" ");
|
||||
|
||||
// Check if this is a g3 binary
|
||||
let is_g3_binary = cmd.get(0).map(|s| s.ends_with("g3")).unwrap_or(false);
|
||||
|
||||
// Check if this is cargo run with g3
|
||||
let is_cargo_run = cmd.get(0).map(|s| s.contains("cargo")).unwrap_or(false)
|
||||
&& cmd.iter().any(|s| s == "run" || s.contains("g3"));
|
||||
|
||||
if !is_g3_binary && !is_cargo_run {
|
||||
return None;
|
||||
}
|
||||
|
||||
// Extract workspace directory
|
||||
let workspace = self.extract_workspace(pid, process, cmd)?;
|
||||
|
||||
// Determine execution method
|
||||
let execution_method = if is_cargo_run {
|
||||
ExecutionMethod::CargoRun
|
||||
} else {
|
||||
ExecutionMethod::Binary
|
||||
};
|
||||
|
||||
// Determine instance type (ensemble if --autonomous flag present)
|
||||
let instance_type = if cmd.iter().any(|s| s == "--autonomous") {
|
||||
InstanceType::Ensemble
|
||||
} else {
|
||||
InstanceType::Single
|
||||
};
|
||||
|
||||
// Extract provider and model
|
||||
let provider = self.extract_flag_value(cmd, "--provider");
|
||||
let model = self.extract_flag_value(cmd, "--model");
|
||||
|
||||
// Get start time
|
||||
let start_time = DateTime::from_timestamp(process.start_time() as i64, 0)
|
||||
.unwrap_or_else(Utc::now);
|
||||
|
||||
// Generate instance ID from PID and start time
|
||||
let id = format!("{}_{}", pid, start_time.timestamp());
|
||||
|
||||
Some(Instance {
|
||||
id,
|
||||
pid: pid.as_u32(),
|
||||
workspace,
|
||||
start_time,
|
||||
status: InstanceStatus::Running,
|
||||
instance_type,
|
||||
provider,
|
||||
model,
|
||||
execution_method,
|
||||
command_line: cmd_str,
|
||||
launch_params: None, // Not available for detected processes
|
||||
})
|
||||
}
|
||||
|
||||
fn extract_workspace(&self, pid: Pid, process: &Process, cmd: &[String]) -> Option<PathBuf> {
|
||||
// Look for --workspace flag
|
||||
for i in 0..cmd.len() {
|
||||
if cmd[i] == "--workspace" && i + 1 < cmd.len() {
|
||||
return Some(PathBuf::from(&cmd[i + 1]));
|
||||
}
|
||||
if cmd[i] == "-w" && i + 1 < cmd.len() {
|
||||
return Some(PathBuf::from(&cmd[i + 1]));
|
||||
}
|
||||
}
|
||||
|
||||
// Fallback: Try to get the working directory of the process
|
||||
#[cfg(target_os = "linux")]
|
||||
{
|
||||
// On Linux, read /proc/<pid>/cwd symlink
|
||||
let cwd_path = format!("/proc/{}/cwd", pid.as_u32());
|
||||
if let Ok(cwd) = std::fs::read_link(&cwd_path) {
|
||||
debug!("Found workspace via /proc for PID {}: {:?}", pid, cwd);
|
||||
return Some(cwd);
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(target_os = "macos")]
|
||||
{
|
||||
// On macOS, use lsof to get the current working directory
|
||||
if let Ok(output) = std::process::Command::new("lsof")
|
||||
.args(["-p", &pid.as_u32().to_string(), "-a", "-d", "cwd", "-Fn"])
|
||||
.output()
|
||||
{
|
||||
if let Ok(stdout) = String::from_utf8(output.stdout) {
|
||||
if let Some(line) = stdout.lines().find(|l| l.starts_with('n')) {
|
||||
let cwd = PathBuf::from(&line[1..]);
|
||||
debug!("Found workspace via lsof for PID {}: {:?}", pid, cwd);
|
||||
return Some(cwd);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Final fallback: use current directory of console
|
||||
warn!("Could not determine workspace for PID {}, using current directory", pid);
|
||||
std::env::current_dir().ok()
|
||||
}
|
||||
|
||||
fn extract_flag_value(&self, cmd: &[String], flag: &str) -> Option<String> {
|
||||
for i in 0..cmd.len() {
|
||||
if cmd[i] == flag && i + 1 < cmd.len() {
|
||||
return Some(cmd[i + 1].clone());
|
||||
}
|
||||
}
|
||||
None
|
||||
}
|
||||
|
||||
pub fn get_process_status(&mut self, pid: u32) -> Option<InstanceStatus> {
|
||||
self.system.refresh_processes();
|
||||
|
||||
let sysinfo_pid = Pid::from_u32(pid);
|
||||
if self.system.process(sysinfo_pid).is_some() {
|
||||
Some(InstanceStatus::Running)
|
||||
} else {
|
||||
Some(InstanceStatus::Terminated)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl Default for ProcessDetector {
|
||||
fn default() -> Self {
|
||||
Self::new()
|
||||
}
|
||||
}
|
||||
5
crates/g3-console/src/process/mod.rs
Normal file
5
crates/g3-console/src/process/mod.rs
Normal file
@@ -0,0 +1,5 @@
|
||||
pub mod detector;
|
||||
pub mod controller;
|
||||
|
||||
pub use detector::*;
|
||||
pub use controller::*;
|
||||
Reference in New Issue
Block a user