Full rewrite

.github/workflows/audit.yml

@@ -12,6 +12,6 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v3
-      - uses: actions-rs/audit-check@v1
+      - uses: rustsec/audit-check@v1.4.1
         with:
           token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/ci.yml

@@ -48,6 +48,23 @@ jobs:
       - name: Run cargo test
         run: cargo test -- --test-threads 1
 
+  msrv:
+    name: Minimum Supported Rust Version
+    runs-on: ubuntu-latest
+    env:
+      RUSTFLAGS: "-D warnings"
+    steps:
+      - name: Checkout sources
+        uses: actions/checkout@v3
+
+      - name: Install MSRV toolchain
+        uses: dtolnay/rust-toolchain@1.63.0
+
+      #- uses: Swatinem/rust-cache@v1
+
+      - name: Run cargo build
+        run: cargo build
+
   lints:
     name: Lints
     runs-on: ubuntu-latest
@@ -115,7 +132,7 @@ jobs:
     runs-on: ubuntu-latest
     environment: production
     if: github.event_name == 'release'
-    needs: [build, test, lints, docs, leaks]
+    needs: [build, test, msrv, lints, docs, leaks]
     steps:
       - name: Checkout sources
         uses: actions/checkout@v3

Cargo.toml

@@ -1,8 +1,9 @@
 [package]
 name = "tokio-graceful-shutdown"
 authors = ["Finomnis <finomnis@gmail.com>"]
-version = "0.13.0"
-edition = "2018"
+version = "0.14.0"
+edition = "2021"
+rust-version = "1.63"
 license = "MIT OR Apache-2.0"
 readme = "README.md"
 repository = "https://github.com/Finomnis/tokio-graceful-shutdown"
@@ -20,39 +21,35 @@ exclude = [
 ]
 
 [dependencies]
-# Error definitions
-thiserror = "1.0.32"
-miette = "5.3.0"
+tracing = { version = "0.1.37", default-features = false }
 
-# For async utilities
-tokio = { version = "1.20.1", default-features = false, features = [
+tokio = { version = "1.32.0", default-features = false, features = [
     "signal",
     "rt",
     "macros",
     "time",
 ] }
-tokio-util = { version = "0.7.2", default-features = false }
-futures = "0.3.23"
-async-recursion = "1.0.0"
-pin-project-lite = "0.2.9"
+tokio-util = { version = "0.7.8", default-features = false }
 
-# For 'IntoSubsystem' trait
-async-trait = "0.1.57"
-
-# For logging
-log = "0.4.17"
+pin-project-lite = "0.2.13"
+thiserror = "1.0.49"
+miette = "5.10.0"
+async-trait = "0.1.73"
+atomic = "0.6.0"
+bytemuck = { version = "1.14.0", features = ["derive"] }
 
 [dev-dependencies]
 # Error propagation
-anyhow = "1.0.61"
+anyhow = "1.0.75"
 eyre = "0.6.8"
-miette = { version = "5.3.0", features = ["fancy"] }
+miette = { version = "5.10.0", features = ["fancy"] }
 
 # Logging
-env_logger = "0.10.0"
+tracing-subscriber = "0.3.17"
+tracing-test = "0.2.4"
 
 # Tokio
-tokio = { version = "1.20.1", features = ["full"] }
+tokio = { version = "1.32.0", features = ["full"] }
 
 # Hyper example
 hyper = { version = "0.14.20", features = ["full"] }

README.md

@@ -40,17 +40,19 @@ This subsystem can now be executed like this:
 ```rust
 #[tokio::main]
 async fn main() -> Result<()> {
-    Toplevel::new()
-        .start("Subsys1", subsys1)
-        .catch_signals()
-        .handle_shutdown_requests(Duration::from_millis(1000))
-        .await
-        .map_err(Into::into)
+    Toplevel::new(|s| async move {
+        s.start(SubsystemBuilder::new("Subsys1", subsys1))
+    })
+    .catch_signals()
+    .handle_shutdown_requests(Duration::from_millis(1000))
+    .await
+    .map_err(Into::into)
 }
 ```
 
 The `Toplevel` object is the root object of the subsystem tree.
-Subsystems can then be started using the `start()` functionality of the toplevel object.
+Subsystems can then be started in it using the `start()` method
+of its `SubsystemHandle` object.
 
 The `catch_signals()` method signals the `Toplevel` object to listen for SIGINT/SIGTERM/Ctrl+C and initiate a shutdown thereafter.
 

TODO.txt

@@ -0,0 +1,29 @@
+- Search for all TODOs in code
+- Port over documentation
+- Port over tests
+
+
+
+
+
+
+Done:
+
+- Name
+- Error handling
+
+- SubsystemBuilder (or PreparedSubsystem, or something similar)!
+    - Allows creating subsystems:
+        - from FnOnce (non-restartable)
+        - from Fn (restartable)
+        - from `trait Subsystem`
+        - from `trait RestartableSubsystem`
+    - maybe gets passed directly into `start`
+
+- Solution to the entire "restartable subsystems" problem:
+    - Make single subsystems awaitable! (Through the object returned from `start`)
+    - Restart is then trivial to implement.
+        - Make subsystem "Shutdown on Error/Panic"
+        - Start/await it in a loop in the parent subsystem
+
+- Error generic

Thoughts.txt

@@ -0,0 +1,48 @@
+need:
+- function that runs on all exit paths of the subsystem
+    - that also has access to the locked parent, to remove itself from the list of children
+- is list of children really important?
+    - atomic counter could work
+    - but what about error propagation?
+
+- error propagation maybe not necessary.
+    - register closure that will be executed on error/shutdown of the child
+    - every subsystem can have 'shutdown triggers' attached to it
+
+fixed facts:
+- subsystems will never change their parent
+
+Suggestions:
+Subsystem have **no** reference to their parents.
+They
+    - Use `joiner_tokens` for awaiting children
+    - Use `cancellation_token` for shutdown
+    - Simply 'drop' a subsystem to hard cancel an entire tree
+
+Open question: How to propagate errors?
+    - Not at all? (Do we need error propagation?)
+    - Through the joiner_tokens?
+        If through the joiner_tokens:
+            - Have `none`/`some(vec)` in every joiner_token for collecting errors
+            - While walking up, put the error in the first available location
+            - When dropping the token, propagate errors up if unconsumed
+            - Uncaught errors simply get printed
+        This should provide a quite natural way of propagating/dropping errors,
+        and should work well with partial shutdown.
+
+Open question: How to deal with errors?
+    - Every spawned subsystem can register functions that handle errors of their children
+    - Possibilities are:
+        - pass further up
+        - Ignore
+        - shutdown self and children
+    - Probably a mechanism that will be inside of the joiner_token
+
+Open question: Ownership?
+    - Parents should own children
+    - But: HOW do children remove themselves from the parent once they are finished?
+        - IMPORTANT QUESTION this is. Might be the one stone that breaks this construct.
+        - The solution *might* be: The joiner_token removes them. That way, there is no recursive dependency.
+        - Might need a fancy data structure that allows efficient RAII based object tracking
+            - solved! (?)
+    -> Implemented in remote_drop_collection

examples/01_normal_shutdown.rs

@@ -7,40 +7,42 @@
 //! If custom arguments for the subsystem coroutines are required,
 //! a struct has to be used instead, as seen in other examples.
 
-use env_logger::{Builder, Env};
 use miette::Result;
 use tokio::time::{sleep, Duration};
-use tokio_graceful_shutdown::{SubsystemHandle, Toplevel};
+use tokio_graceful_shutdown::{SubsystemBuilder, SubsystemHandle, Toplevel};
 
 async fn subsys1(subsys: SubsystemHandle) -> Result<()> {
-    log::info!("Subsystem1 started.");
+    tracing::info!("Subsystem1 started.");
     subsys.on_shutdown_requested().await;
-    log::info!("Shutting down Subsystem1 ...");
+    tracing::info!("Shutting down Subsystem1 ...");
     sleep(Duration::from_millis(400)).await;
-    log::info!("Subsystem1 stopped.");
+    tracing::info!("Subsystem1 stopped.");
     Ok(())
 }
 
 async fn subsys2(subsys: SubsystemHandle) -> Result<()> {
-    log::info!("Subsystem2 started.");
+    tracing::info!("Subsystem2 started.");
     subsys.on_shutdown_requested().await;
-    log::info!("Shutting down Subsystem2 ...");
+    tracing::info!("Shutting down Subsystem2 ...");
     sleep(Duration::from_millis(500)).await;
-    log::info!("Subsystem2 stopped.");
+    tracing::info!("Subsystem2 stopped.");
     Ok(())
 }
 
 #[tokio::main]
 async fn main() -> Result<()> {
     // Init logging
-    Builder::from_env(Env::default().default_filter_or("debug")).init();
+    tracing_subscriber::fmt()
+        .with_max_level(tracing::Level::TRACE)
+        .init();
 
-    // Create toplevel
-    Toplevel::new()
-        .start("Subsys1", subsys1)
-        .start("Subsys2", subsys2)
-        .catch_signals()
-        .handle_shutdown_requests(Duration::from_millis(1000))
-        .await
-        .map_err(Into::into)
+    // Setup and execute subsystem tree
+    Toplevel::new(|s| async move {
+        s.start(SubsystemBuilder::new("Subsys1", subsys1));
+        s.start(SubsystemBuilder::new("Subsys2", subsys2));
+    })
+    .catch_signals()
+    .handle_shutdown_requests(Duration::from_millis(1000))
+    .await
+    .map_err(Into::into)
 }

examples/02_structs.rs

@@ -2,27 +2,26 @@
 //! custom parameters to be passed to the subsystem.
 //!
 //! There are two ways of using structs as subsystems, by either
-//! wrapping them in an async closure, or by implementing the
+//! wrapping them in a closure, or by implementing the
 //! IntoSubsystem trait. Note, though, that the IntoSubsystem
 //! trait requires an additional dependency, `async-trait`.
 
 use async_trait::async_trait;
-use env_logger::{Builder, Env};
 use miette::Result;
 use tokio::time::{sleep, Duration};
-use tokio_graceful_shutdown::{IntoSubsystem, SubsystemHandle, Toplevel};
+use tokio_graceful_shutdown::{IntoSubsystem, SubsystemBuilder, SubsystemHandle, Toplevel};
 
 struct Subsystem1 {
     arg: u32,
 }
 
 impl Subsystem1 {
     async fn run(self, subsys: SubsystemHandle) -> Result<()> {
-        log::info!("Subsystem1 started. Extra argument: {}", self.arg);
+        tracing::info!("Subsystem1 started. Extra argument: {}", self.arg);
         subsys.on_shutdown_requested().await;
-        log::info!("Shutting down Subsystem1 ...");
+        tracing::info!("Shutting down Subsystem1 ...");
         sleep(Duration::from_millis(500)).await;
-        log::info!("Subsystem1 stopped.");
+        tracing::info!("Subsystem1 stopped.");
         Ok(())
     }
 }
@@ -34,29 +33,32 @@ struct Subsystem2 {
 #[async_trait]
 impl IntoSubsystem<miette::Report> for Subsystem2 {
     async fn run(self, subsys: SubsystemHandle) -> Result<()> {
-        log::info!("Subsystem2 started. Extra argument: {}", self.arg);
+        tracing::info!("Subsystem2 started. Extra argument: {}", self.arg);
         subsys.on_shutdown_requested().await;
-        log::info!("Shutting down Subsystem2 ...");
+        tracing::info!("Shutting down Subsystem2 ...");
         sleep(Duration::from_millis(500)).await;
-        log::info!("Subsystem2 stopped.");
+        tracing::info!("Subsystem2 stopped.");
         Ok(())
     }
 }
 
 #[tokio::main]
 async fn main() -> Result<()> {
     // Init logging
-    Builder::from_env(Env::default().default_filter_or("debug")).init();
+    tracing_subscriber::fmt()
+        .with_max_level(tracing::Level::TRACE)
+        .init();
 
     let subsys1 = Subsystem1 { arg: 42 };
     let subsys2 = Subsystem2 { arg: 69 };
 
-    // Create toplevel
-    Toplevel::new()
-        .start("Subsys1", |a| subsys1.run(a))
-        .start("Subsys2", subsys2.into_subsystem())
-        .catch_signals()
-        .handle_shutdown_requests(Duration::from_millis(1000))
-        .await
-        .map_err(Into::into)
+    // Setup and execute subsystem tree
+    Toplevel::new(|s| async move {
+        s.start(SubsystemBuilder::new("Subsys1", |a| subsys1.run(a)));
+        s.start(SubsystemBuilder::new("Subsys2", subsys2.into_subsystem()));
+    })
+    .catch_signals()
+    .handle_shutdown_requests(Duration::from_millis(1000))
+    .await
+    .map_err(Into::into)
 }

examples/03_shutdown_timeout.rs

@@ -4,30 +4,32 @@
 //! so the subsystem gets cancelled and the program returns an appropriate
 //! error code.
 
-use env_logger::{Builder, Env};
 use miette::Result;
 use tokio::time::{sleep, Duration};
-use tokio_graceful_shutdown::{SubsystemHandle, Toplevel};
+use tokio_graceful_shutdown::{SubsystemBuilder, SubsystemHandle, Toplevel};
 
 async fn subsys1(subsys: SubsystemHandle) -> Result<()> {
-    log::info!("Subsystem1 started.");
+    tracing::info!("Subsystem1 started.");
     subsys.on_shutdown_requested().await;
-    log::info!("Shutting down Subsystem1 ...");
+    tracing::info!("Shutting down Subsystem1 ...");
     sleep(Duration::from_millis(2000)).await;
-    log::info!("Subsystem1 stopped.");
+    tracing::info!("Subsystem1 stopped.");
     Ok(())
 }
 
 #[tokio::main]
 async fn main() -> Result<()> {
     // Init logging
-    Builder::from_env(Env::default().default_filter_or("debug")).init();
+    tracing_subscriber::fmt()
+        .with_max_level(tracing::Level::TRACE)
+        .init();
 
-    // Create toplevel
-    Toplevel::new()
-        .start("Subsys1", subsys1)
-        .catch_signals()
-        .handle_shutdown_requests(Duration::from_millis(500))
-        .await
-        .map_err(Into::into)
+    // Setup and execute subsystem tree
+    Toplevel::new(|s| async move {
+        s.start(SubsystemBuilder::new("Subsys1", subsys1));
+    })
+    .catch_signals()
+    .handle_shutdown_requests(Duration::from_millis(500))
+    .await
+    .map_err(Into::into)
 }