gitlore/tests/http_parity_tests.rs

//! HTTP behavior parity tests: verify asupersync h1 matches reqwest semantics
//! that `lore` depends on.
//!
//! These tests confirm six critical behaviors:
//! 1. Auto redirect: 301 -> follows Location header
//! 2. Proxy: HTTP_PROXY not supported (documented)
//! 3. Connection keep-alive: sequential requests reuse connections
//! 4. System DNS: hostname resolution works
//! 5. Content-Length on POST: header is auto-added
//! 6. TLS cert validation: invalid certs are rejected

use std::io::{Read, Write};
use std::net::TcpListener;
use std::sync::Arc;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::time::Duration;

use lore::http::Client;

/// Run an async block on the asupersync runtime.
fn run<F: std::future::Future<Output = T>, T>(f: F) -> T {
    asupersync::runtime::RuntimeBuilder::new()
        .build()
        .unwrap()
        .block_on(f)
}

/// Read one HTTP request from a stream (drain until double-CRLF), returning
/// the raw request bytes (headers only).
fn drain_request(stream: &mut std::net::TcpStream) -> Vec<u8> {
    let mut buf = Vec::new();
    let mut tmp = [0u8; 1];
    loop {
        let n = stream.read(&mut tmp).unwrap();
        if n == 0 {
            break;
        }
        buf.extend_from_slice(&tmp[..n]);
        if buf.len() >= 4 && buf[buf.len() - 4..] == *b"\r\n\r\n" {
            break;
        }
    }
    buf
}

// =========================================================================
// Test 1: Auto redirect — 301 is followed transparently
// =========================================================================

#[test]
fn redirect_301_is_followed() {
    let listener = TcpListener::bind("127.0.0.1:0").unwrap();
    let port = listener.local_addr().unwrap().port();

    std::thread::spawn(move || {
        // First request: return 301 with Location pointing to /final
        let (mut stream, _) = listener.accept().unwrap();
        drain_request(&mut stream);

        let redirect = format!(
            "HTTP/1.1 301 Moved Permanently\r\n\
             Location: http://127.0.0.1:{port}/final\r\n\
             Content-Length: 0\r\n\
             \r\n"
        );
        stream.write_all(redirect.as_bytes()).unwrap();
        stream.flush().unwrap();
        drop(stream);

        // Second request (after redirect): return 200 with body
        let (mut stream2, _) = listener.accept().unwrap();
        drain_request(&mut stream2);

        let body = r#"{"redirected":true}"#;
        let response = format!(
            "HTTP/1.1 200 OK\r\n\
             Content-Type: application/json\r\n\
             Content-Length: {}\r\n\
             \r\n\
             {body}",
            body.len()
        );
        stream2.write_all(response.as_bytes()).unwrap();
        stream2.flush().unwrap();
    });

    let base = format!("http://127.0.0.1:{port}");

    run(async {
        let client = Client::with_timeout(Duration::from_secs(5));
        let resp = client.get(&format!("{base}/original"), &[]).await.unwrap();

        assert!(
            resp.is_success(),
            "expected 200 after redirect, got {}",
            resp.status
        );
        assert_eq!(resp.status, 200);

        let parsed: serde_json::Value = resp.json().unwrap();
        assert_eq!(parsed["redirected"], true);
    });
}

// =========================================================================
// Test 2: Proxy — HTTP_PROXY is NOT auto-detected (documented difference)
// =========================================================================

#[test]
fn proxy_env_not_auto_detected() {
    // Set HTTP_PROXY to a bogus address. If the client respected it, the
    // request would fail connecting to the proxy. Since asupersync ignores
    // proxy env vars, the request should go directly to the target.
    let body = r#"{"direct":true}"#;
    let listener = TcpListener::bind("127.0.0.1:0").unwrap();
    let port = listener.local_addr().unwrap().port();

    std::thread::spawn(move || {
        let (mut stream, _) = listener.accept().unwrap();
        drain_request(&mut stream);

        let response = format!(
            "HTTP/1.1 200 OK\r\n\
             Content-Type: application/json\r\n\
             Content-Length: {}\r\n\
             \r\n\
             {body}",
            body.len()
        );
        stream.write_all(response.as_bytes()).unwrap();
        stream.flush().unwrap();
    });

    // Set a bogus proxy — the client should ignore it.
    // SAFETY: test-only; no other thread reads HTTP_PROXY concurrently.
    unsafe { std::env::set_var("HTTP_PROXY", "http://192.0.2.1:9999") };

    let base = format!("http://127.0.0.1:{port}");

    run(async {
        let client = Client::with_timeout(Duration::from_secs(5));
        let resp = client.get(&format!("{base}/api/test"), &[]).await.unwrap();

        assert!(resp.is_success());
        let parsed: serde_json::Value = resp.json().unwrap();
        assert_eq!(parsed["direct"], true);
    });

    // Clean up env var.
    // SAFETY: test-only; no other thread reads HTTP_PROXY concurrently.
    unsafe { std::env::remove_var("HTTP_PROXY") };
}

// =========================================================================
// Test 3: Connection keep-alive — sequential requests to same host
// =========================================================================

#[test]
fn sequential_requests_connect_separately() {
    // Track how many TCP connections are accepted. Each request should
    // establish its own connection (current behavior — pool not yet wired).
    let connection_count = Arc::new(AtomicUsize::new(0));
    let count_clone = Arc::clone(&connection_count);

    let listener = TcpListener::bind("127.0.0.1:0").unwrap();
    let port = listener.local_addr().unwrap().port();

    std::thread::spawn(move || {
        for _ in 0..3 {
            let (mut stream, _) = listener.accept().unwrap();
            count_clone.fetch_add(1, Ordering::SeqCst);
            drain_request(&mut stream);

            let body = r#"{"ok":true}"#;
            let response = format!(
                "HTTP/1.1 200 OK\r\n\
                 Content-Type: application/json\r\n\
                 Content-Length: {}\r\n\
                 Connection: keep-alive\r\n\
                 \r\n\
                 {body}",
                body.len()
            );
            stream.write_all(response.as_bytes()).unwrap();
            stream.flush().unwrap();
        }
    });

    let base = format!("http://127.0.0.1:{port}");

    run(async {
        let client = Client::with_timeout(Duration::from_secs(5));

        for _ in 0..3 {
            let resp = client.get(&format!("{base}/api/data"), &[]).await.unwrap();
            assert!(resp.is_success());
        }
    });

    let total = connection_count.load(Ordering::SeqCst);
    // Document current behavior: each request opens a new connection.
    // If/when connection pooling is wired, this assertion should change
    // to assert!(total <= 2) to verify keep-alive.
    assert!(
        (1..=3).contains(&total),
        "expected 1-3 connections (got {total}); \
         3 = no pooling, 1 = full keep-alive"
    );
}

// =========================================================================
// Test 4: System DNS — localhost resolves and connects
// =========================================================================

#[test]
fn system_dns_resolves_localhost() {
    let body = r#"{"dns":"ok"}"#;
    let listener = TcpListener::bind("127.0.0.1:0").unwrap();
    let port = listener.local_addr().unwrap().port();

    std::thread::spawn(move || {
        let (mut stream, _) = listener.accept().unwrap();
        drain_request(&mut stream);

        let response = format!(
            "HTTP/1.1 200 OK\r\n\
             Content-Type: application/json\r\n\
             Content-Length: {}\r\n\
             \r\n\
             {body}",
            body.len()
        );
        stream.write_all(response.as_bytes()).unwrap();
        stream.flush().unwrap();
    });

    run(async {
        let client = Client::with_timeout(Duration::from_secs(5));
        // Use "localhost" instead of "127.0.0.1" to exercise DNS resolution.
        let resp = client
            .get(
                &format!("http://localhost:{port}/api/dns-test"),
                &[("Accept", "application/json")],
            )
            .await
            .unwrap();

        assert!(resp.is_success());
        let parsed: serde_json::Value = resp.json().unwrap();
        assert_eq!(parsed["dns"], "ok");
    });
}

// =========================================================================
// Test 5: Content-Length on POST — header is auto-added by codec
// =========================================================================

#[test]
fn post_includes_content_length_header() {
    let listener = TcpListener::bind("127.0.0.1:0").unwrap();
    let port = listener.local_addr().unwrap().port();

    let (tx, rx) = std::sync::mpsc::channel::<String>();

    std::thread::spawn(move || {
        let (mut stream, _) = listener.accept().unwrap();
        let request_bytes = drain_request(&mut stream);
        let request_text = String::from_utf8_lossy(&request_bytes).to_string();
        tx.send(request_text).unwrap();

        let body = r#"{"received":true}"#;
        let response = format!(
            "HTTP/1.1 200 OK\r\n\
             Content-Type: application/json\r\n\
             Content-Length: {}\r\n\
             \r\n\
             {body}",
            body.len()
        );
        stream.write_all(response.as_bytes()).unwrap();
        stream.flush().unwrap();
    });

    let base = format!("http://127.0.0.1:{port}");

    run(async {
        let client = Client::with_timeout(Duration::from_secs(5));

        #[derive(serde::Serialize)]
        struct Payload {
            model: String,
            input: Vec<String>,
        }

        let payload = Payload {
            model: "test-model".into(),
            input: vec!["hello".into()],
        };

        let resp = client
            .post_json(&format!("{base}/api/embed"), &[], &payload)
            .await
            .unwrap();

        assert!(resp.is_success());
    });

    let captured = rx.recv_timeout(Duration::from_secs(5)).unwrap();

    // Verify Content-Length header was present in the request.
    let has_content_length = captured
        .lines()
        .any(|line| line.to_lowercase().starts_with("content-length:"));
    assert!(
        has_content_length,
        "POST request should include Content-Length header.\n\
         Captured request:\n{captured}"
    );

    // Verify Content-Length value matches actual body length.
    let cl_value: usize = captured
        .lines()
        .find(|line| line.to_lowercase().starts_with("content-length:"))
        .and_then(|line| line.split(':').nth(1))
        .and_then(|v| v.trim().parse().ok())
        .expect("Content-Length should be a valid number");
    assert!(
        cl_value > 0,
        "Content-Length should be > 0 for non-empty POST body"
    );
}

// =========================================================================
// Test 6: TLS cert validation — self-signed/invalid cert is rejected
// =========================================================================

#[test]
fn tls_rejects_plain_tcp_as_https() {
    // Start a plain TCP server and try to connect via https://.
    // The TLS handshake should fail because the server doesn't speak TLS.
    let listener = TcpListener::bind("127.0.0.1:0").unwrap();
    let port = listener.local_addr().unwrap().port();

    std::thread::spawn(move || {
        // Accept and hold the connection so the client can attempt TLS.
        if let Ok((mut stream, _)) = listener.accept() {
            // Send garbage — the TLS handshake will fail on the client side.
            let _ = stream.write_all(b"NOT TLS\r\n");
            std::thread::sleep(Duration::from_secs(2));
        }
    });

    run(async {
        let client = Client::with_timeout(Duration::from_secs(5));
        let result = client
            .get(&format!("https://127.0.0.1:{port}/api/test"), &[])
            .await;

        assert!(
            result.is_err(),
            "expected TLS error when connecting to plain TCP"
        );
        let err_str = format!("{:?}", result.unwrap_err());
        // The error should be TLS-related (not a generic connection error).
        assert!(
            err_str.contains("Tls")
                || err_str.to_lowercase().contains("tls")
                || err_str.to_lowercase().contains("ssl")
                || err_str.to_lowercase().contains("handshake")
                || err_str.to_lowercase().contains("certificate"),
            "error should mention TLS/SSL, got: {err_str}"
        );
    });
}

// =========================================================================
// Test 6b: TLS cert validation — connection to unreachable host fails
// =========================================================================

#[test]
fn tls_connection_to_nonexistent_host_fails() {
    run(async {
        let client = Client::with_timeout(Duration::from_secs(3));
        // 192.0.2.1 is TEST-NET-1 (RFC 5737) — guaranteed unroutable.
        let result = client.get("https://192.0.2.1:443/api/test", &[]).await;

        assert!(
            result.is_err(),
            "expected error connecting to unroutable host"
        );
    });
}