release: ライブラリ更新対応

Cargo.toml

@@ -24,17 +24,23 @@ members = [
 [workspace.dependencies]
 anyhow = "1.0.82"
 async-graphql = { version = "7.0.9" }
+bytes = "1"
 convert_case = "0.6.0"
 derive_more = { version = "1", features = ["from", "debug"] }
 futures-util = { version = "0.3.30" }
-headers = "0.3.9" # previous version until hyper is updated to 1+
-http = "0.2.12" # previous version until hyper is updated to 1+
+headers = "0.4"
+http = "1.0"
+http-body = "1.0"
+http-body-util = "0.1"
+hyper = { version = "1.0", features = ["server", "http1", "http2"] }
+hyper-util = { version = "0.1", features = ["server", "tokio", "http1", "http2"] }
 indexmap = "2.2.6"
 insta = { version = "1.38.0", features = ["json"] }
 lazy_static = "1.4.0"
-reqwest = { version = "0.11", features = [
+reqwest = { version = "0.12", features = [
   "json",
   "rustls-tls",
+  "http2",
 ], default-features = false }
 serde = { version = "1.0.200", features = ["derive"] }
 serde_json = { version = "1.0.116", features = ["preserve_order"] }
@@ -65,6 +71,7 @@ cli = [
   "dep:moka",
   "dep:hyper-rustls",
   "dep:rustls",
+  "dep:tokio-rustls",
   "dep:inquire",
   "dep:which",
   "dep:update-informer",
@@ -77,6 +84,7 @@ cli = [
   "dep:tailcall-version",
   "dep:genai",
   "dep:ctrlc",
+  "dep:reqwest-middleware",
 ]
 
 # Feature flag to enable all default features.
@@ -103,6 +111,7 @@ async-std = { version = "1.12.0", features = [
 ] }
 async-trait = "0.1.80"
 base64 = "0.22.1"
+bytes = "1"
 cache_control = "0.2.0"
 chrono = "0.4.38"
 clap = { version = "4.5.4", features = ["derive"] }
@@ -126,11 +135,14 @@ getrandom = { version = "0.2.14", features = ["js"] }
 headers = { workspace = true }
 htpasswd-verify = { version = "0.3.0", git = "https://github.com/twistedfall/htpasswd-verify", rev = "ff14703083cbd639f7d05622b398926f3e718d61" } # fork version that is wasm compatible
 http = { workspace = true }
-http-cache-reqwest = { version = "0.13.0", features = [
+http-body = { workspace = true }
+http-body-util = { workspace = true }
+http-cache-reqwest = { version = "0.14.0", features = [
   "manager-moka",
 ], default-features = false, optional = true }
-hyper = { version = "0.14.28", features = ["server"], default-features = false }
-hyper-rustls = { version = "0.25.0", optional = true }
+hyper = { workspace = true }
+hyper-util = { workspace = true }
+hyper-rustls = { version = "0.27.0", optional = true }
 indenter = "0.3.3"
 indexmap = { workspace = true }
 inquire = { version = "0.7.5", optional = true }
@@ -149,25 +161,26 @@ nom = "7.1.3"
 num = "0.4.3"
 num_cpus = "1.16.0"
 once_cell = "1.19.0"
-opentelemetry = { version = "0.23.0", features = ["trace", "logs", "metrics"] }
-opentelemetry-appender-tracing = { version = "0.4.0" }
-opentelemetry-http = "0.12.0"
-opentelemetry-otlp = { version = "0.16.0", features = [
+opentelemetry = { version = "0.28", features = ["trace", "logs", "metrics"] }
+opentelemetry-appender-tracing = { version = "0.28" }
+opentelemetry-http = "0.28"
+opentelemetry-otlp = { version = "0.28", features = [
   "trace",
   "logs",
   "metrics",
   # required to make grpc requests
+  "grpc-tonic",
   "tls-roots",
 ], optional = true }
-opentelemetry-prometheus = "0.16.0"
-opentelemetry-semantic-conventions = "0.15.0"
-opentelemetry-stdout = { version = "0.4.0", features = [
+opentelemetry-prometheus = "0.28"
+opentelemetry-semantic-conventions = "0.28"
+opentelemetry-stdout = { version = "0.28", features = [
   "trace",
   "logs",
   "metrics",
 ] }
-opentelemetry-system-metrics = { version = "0.2.0", optional = true }
-opentelemetry_sdk = { version = "0.23.0", features = [
+opentelemetry-system-metrics = { version = "0.3", optional = true }
+opentelemetry_sdk = { version = "0.28", features = [
   "trace",
   "logs",
   "metrics",
@@ -184,11 +197,12 @@ protox-parse = "0.7.0"
 rand = "0.8.5"
 regex = "1.10.4"
 reqwest = { workspace = true }
-reqwest-middleware = "0.2.5"
+reqwest-middleware = { version = "0.3", optional = true }
 resource = "0.5.0"
 rquickjs = { "version" = "0.7.0", optional = true, features = ["macro"] }
 rustls = { version = "0.23.5", optional = true, features = [
   "std",
+  "ring",
 ], default-features = false }
 # dependencies safe for wasm:
 
@@ -214,11 +228,12 @@ tailcall-valid = { workspace = true }
 tailcall-version = { path = "./tailcall-version", optional = true }
 thiserror = { workspace = true }
 tokio = { workspace = true }
-tonic = { version = "0.11.0", default-features = false }
-tonic-types = "0.12.1"
+tokio-rustls = { version = "0.26", optional = true }
+tonic = { version = "0.12", default-features = false }
+tonic-types = "0.12"
 # TODO: disable some levels with features?
 tracing = { workspace = true }
-tracing-opentelemetry = "0.24.0"
+tracing-opentelemetry = "0.29"
 tracing-subscriber = { version = "0.3.18", features = [
   "default",
   "fmt",
@@ -243,17 +258,11 @@ rquickjs = { "version" = "0.7.0", optional = true, features = [
 
 [dev-dependencies]
 bincode = "1.3.3"
-cacache = { version = "13.0.0", default-features = false, features = [
-  "tokio-runtime",
-  "mmap",
-] }
+cacache = { version = "13.0.0", features = ["mmap"] }
 criterion = "0.5.1"
 datatest-stable = "0.2.9"
 flate2 = "1.0.30"
-http-cache-semantics = { version = "1.0.1", default-features = false, features = [
-  "with_serde",
-  "reqwest",
-] }
+http-cache-semantics = { version = "2.1.0", features = ["serde"] }
 httpmock = "0.7.0"
 insta = { workspace = true }
 maplit = "1.0.2"

benches/handle_request_bench.rs

@@ -1,7 +1,9 @@
 use std::sync::Arc;
 
+use bytes::Bytes;
 use criterion::Criterion;
 use http::Request;
+use http_body_util::Full;
 use tailcall::cli::server::server_config::ServerConfig;
 use tailcall::core::async_graphql_hyper::GraphQLRequest;
 use tailcall::core::blueprint::Blueprint;
@@ -35,10 +37,10 @@ pub fn benchmark_handle_request(c: &mut Criterion) {
                 let req = Request::builder()
                     .method("POST")
                     .uri("http://localhost:8000/graphql")
-                    .body(hyper::Body::from(QUERY))
+                    .body(Full::new(Bytes::from(QUERY)))
                     .unwrap();
 
-                let _ = handle_request::<GraphQLRequest>(req, server_config.app_ctx.clone())
+                let _ = handle_request::<GraphQLRequest, _>(req, server_config.app_ctx.clone())
                     .await
                     .unwrap();
             });
@@ -58,10 +60,10 @@ pub fn benchmark_handle_request(c: &mut Criterion) {
                 let req = Request::builder()
                     .method("POST")
                     .uri("http://localhost:8000/graphql")
-                    .body(hyper::Body::from(QUERY))
+                    .body(Full::new(Bytes::from(QUERY)))
                     .unwrap();
 
-                let _ = handle_request::<GraphQLRequest>(req, server_config.app_ctx.clone())
+                let _ = handle_request::<GraphQLRequest, _>(req, server_config.app_ctx.clone())
                     .await
                     .unwrap();
             });

src/cli/metrics.rs

@@ -5,29 +5,30 @@ use crate::core::runtime::TargetRuntime;
 fn cache_metrics(runtime: &TargetRuntime) -> Result<()> {
     let meter = opentelemetry::global::meter("cache");
     let cache = runtime.cache.clone();
-    let counter = meter
+    let _counter = meter
         .f64_observable_gauge("cache.hit_rate")
         .with_description("Cache hit rate ratio")
-        .init();
-
-    meter.register_callback(&[counter.as_any()], move |observer| {
-        if let Some(hit_rate) = cache.hit_rate() {
-            observer.observe_f64(&counter, hit_rate, &[]);
-        }
-    })?;
+        .with_callback(move |observer| {
+            if let Some(hit_rate) = cache.hit_rate() {
+                observer.observe(hit_rate, &[]);
+            }
+        })
+        .build();
 
     Ok(())
 }
 
-fn process_resources_metrics() -> Result<()> {
+async fn process_resources_metrics() -> Result<()> {
     let meter = opentelemetry::global::meter("process-resources");
 
-    opentelemetry_system_metrics::init_process_observer(meter).map_err(|err| anyhow!(err))
+    opentelemetry_system_metrics::init_process_observer(meter)
+        .await
+        .map_err(|err| anyhow!(err))
 }
 
-pub fn init_metrics(runtime: &TargetRuntime) -> Result<()> {
+pub async fn init_metrics(runtime: &TargetRuntime) -> Result<()> {
     cache_metrics(runtime)?;
-    process_resources_metrics()?;
+    process_resources_metrics().await?;
 
     Ok(())
 }

src/cli/runtime/http.rs

@@ -27,7 +27,7 @@ static HTTP_CLIENT_REQUEST_COUNT: Lazy<Counter<u64>> = Lazy::new(|| {
     meter
         .u64_counter("http.client.request.count")
         .with_description("Number of outgoing requests")
-        .init()
+        .build()
 });
 
 #[derive(Default)]
@@ -92,9 +92,6 @@ impl NativeHttp {
             .tcp_keepalive(Some(Duration::from_secs(upstream.tcp_keep_alive)))
             .timeout(Duration::from_secs(upstream.timeout))
             .connect_timeout(Duration::from_secs(upstream.connect_timeout))
-            .http2_keep_alive_interval(Some(Duration::from_secs(upstream.keep_alive_interval)))
-            .http2_keep_alive_timeout(Duration::from_secs(upstream.keep_alive_timeout))
-            .http2_keep_alive_while_idle(upstream.keep_alive_while_idle)
             .pool_idle_timeout(Some(Duration::from_secs(upstream.pool_idle_timeout)))
             .pool_max_idle_per_host(upstream.pool_max_idle_per_host)
             .user_agent(upstream.user_agent.clone())

src/cli/server/http_1.rs

@@ -1,38 +1,22 @@
 use std::sync::Arc;
 
-use hyper::service::{make_service_fn, service_fn};
+use hyper::server::conn::http1;
+use hyper::service::service_fn;
+use hyper_util::rt::TokioIo;
+use tokio::net::TcpListener;
 use tokio::sync::oneshot;
 
 use super::server_config::ServerConfig;
 use crate::core::async_graphql_hyper::{GraphQLBatchRequest, GraphQLRequest};
 use crate::core::http::handle_request;
-use crate::core::Errata;
 
 pub async fn start_http_1(
     sc: Arc<ServerConfig>,
     server_up_sender: Option<oneshot::Sender<()>>,
 ) -> anyhow::Result<()> {
     let addr = sc.addr();
-    let make_svc_single_req = make_service_fn(|_conn| {
-        let state = Arc::clone(&sc);
-        async move {
-            Ok::<_, anyhow::Error>(service_fn(move |req| {
-                handle_request::<GraphQLRequest>(req, state.app_ctx.clone())
-            }))
-        }
-    });
-
-    let make_svc_batch_req = make_service_fn(|_conn| {
-        let state = Arc::clone(&sc);
-        async move {
-            Ok::<_, anyhow::Error>(service_fn(move |req| {
-                handle_request::<GraphQLBatchRequest>(req, state.app_ctx.clone())
-            }))
-        }
-    });
-    let builder = hyper::Server::try_bind(&addr)
-        .map_err(Errata::from)?
-        .http1_pipeline_flush(sc.app_ctx.blueprint.server.pipeline_flush);
+    let listener = TcpListener::bind(&addr).await?;
+
     super::log_launch(sc.as_ref());
 
     if let Some(sender) = server_up_sender {
@@ -41,14 +25,32 @@ pub async fn start_http_1(
             .or(Err(anyhow::anyhow!("Failed to send message")))?;
     }
 
-    let server: std::prelude::v1::Result<(), hyper::Error> =
-        if sc.blueprint.server.enable_batch_requests {
-            builder.serve(make_svc_batch_req).await
-        } else {
-            builder.serve(make_svc_single_req).await
-        };
-
-    let result = server.map_err(Errata::from);
-
-    Ok(result?)
+    let enable_batch = sc.app_ctx.blueprint.server.enable_batch_requests;
+    let pipeline_flush = sc.app_ctx.blueprint.server.pipeline_flush;
+
+    loop {
+        let (stream, _) = listener.accept().await?;
+        let io = TokioIo::new(stream);
+        let app_ctx = sc.app_ctx.clone();
+
+        tokio::task::spawn(async move {
+            let service = service_fn(move |req| {
+                let app_ctx = app_ctx.clone();
+                async move {
+                    if enable_batch {
+                        handle_request::<GraphQLBatchRequest, _>(req, app_ctx).await
+                    } else {
+                        handle_request::<GraphQLRequest, _>(req, app_ctx).await
+                    }
+                }
+            });
+
+            let mut conn = http1::Builder::new();
+            conn.pipeline_flush(pipeline_flush);
+
+            if let Err(err) = conn.serve_connection(io, service).await {
+                tracing::debug!("Error serving connection: {:?}", err);
+            }
+        });
+    }
 }