use crate::{ BuildArtifacts, BuildRequest, BuildStage, BuilderUpdate, Platform, ProgressRx, ProgressTx, Result, StructuredOutput, }; use anyhow::Context; use dioxus_cli_opt::process_file_to; use futures_util::{future::OptionFuture, pin_mut, FutureExt}; use std::{ env, time::{Duration, Instant, SystemTime}, }; use std::{ net::SocketAddr, path::{Path, PathBuf}, process::Stdio, }; use subsecond_types::JumpTable; use tokio::{ io::{AsyncBufReadExt, BufReader, Lines}, process::{Child, ChildStderr, ChildStdout, Command}, task::JoinHandle, }; use super::{BuildContext, BuildId, BuildMode, HotpatchModuleCache}; /// The component of the serve engine that watches ongoing builds and manages their state, open handle, /// and progress. /// /// Previously, the builder allowed multiple apps to be built simultaneously, but this newer design /// simplifies the code and allows only one app and its server to be built at a time. /// /// Here, we track the number of crates being compiled, assets copied, the times of these events, and /// other metadata that gives us useful indicators for the UI. /// /// A handle to a running app. /// /// The actual child processes might not be present (web) or running (died/killed). /// /// The purpose of this struct is to accumulate state about the running app and its server, like /// any runtime information needed to hotreload the app or send it messages. /// /// We might want to bring in websockets here too, so we know the exact channels the app is using to /// communicate with the devserver. Currently that's a broadcast-type system, so this struct isn't super /// duper useful. /// /// todo: restructure this such that "open" is a running task instead of blocking the main thread pub(crate) struct AppBuilder { pub tx: ProgressTx, pub rx: ProgressRx, // The original request with access to its build directory pub build: BuildRequest, // Ongoing build task, if any pub build_task: JoinHandle>, // If a build has already finished, we'll have its artifacts (rustc, link args, etc) to work with pub artifacts: Option, /// The aslr offset of this running app pub aslr_reference: Option, /// The list of patches applied to the app, used to know which ones to reapply and/or iterate from. pub patches: Vec, pub patch_cache: Option, /// The virtual directory that assets will be served from /// Used mostly for apk/ipa builds since they live in simulator pub runtime_asset_dir: Option, // These might be None if the app died or the user did not specify a server pub child: Option, // stdio for the app so we can read its stdout/stderr // we don't map stdin today (todo) but most apps don't need it pub stdout: Option>>, pub stderr: Option>>, /// The executables but with some extra entropy in their name so we can run two instances of the /// same app without causing collisions on the filesystem. pub entropy_app_exe: Option, pub builds_opened: usize, // Metadata about the build that needs to be managed by watching build updates // used to render the TUI pub stage: BuildStage, pub compiled_crates: usize, pub expected_crates: usize, pub bundling_progress: f64, pub compile_start: Option, pub compile_end: Option, pub bundle_start: Option, pub bundle_end: Option, } impl AppBuilder { /// Create a new `AppBuilder` and immediately start a build process. /// /// This method initializes the builder with the provided `BuildRequest` and spawns an asynchronous /// task (`build_task`) to handle the build process. The build process involves several stages: /// /// 1. **Tooling Verification**: Ensures that the necessary tools are available for the build. /// 2. **Build Directory Preparation**: Sets up the directory structure required for the build. /// 3. **Build Execution**: Executes the build process asynchronously. /// 4. **Bundling**: Packages the built artifacts into a final bundle. /// /// The `build_task` is a Tokio task that runs the build process in the background. It uses a /// `BuildContext` to manage the build state and communicate progress or errors via a message /// channel (`tx`). /// /// The builder is initialized with default values for various fields, such as the build stage, /// progress metrics, and optional runtime configurations. /// /// # Notes /// /// - The `build_task` is immediately spawned and will run independently of the caller. /// - The caller can use other methods on the `AppBuilder` to monitor the build progress or handle /// updates (e.g., `wait`, `finish_build`). /// - The build process is designed to be cancellable and restartable using methods like `abort_all` /// or `rebuild`. pub(crate) fn start(request: &BuildRequest, mode: BuildMode) -> Result { let (tx, rx) = futures_channel::mpsc::unbounded(); Ok(Self { build: request.clone(), stage: BuildStage::Initializing, build_task: tokio::spawn({ let request = request.clone(); let tx = tx.clone(); async move { let ctx = BuildContext { mode, tx: tx.clone(), }; request.verify_tooling(&ctx).await?; request.prepare_build_dir()?; request.build(&ctx).await } }), tx, rx, patches: vec![], compiled_crates: 0, expected_crates: 1, bundling_progress: 0.0, builds_opened: 0, compile_start: Some(Instant::now()), aslr_reference: None, compile_end: None, bundle_start: None, bundle_end: None, runtime_asset_dir: None, child: None, stderr: None, stdout: None, entropy_app_exe: None, artifacts: None, patch_cache: None, }) } /// Wait for any new updates to the builder - either it completed or gave us a message etc pub(crate) async fn wait(&mut self) -> BuilderUpdate { use futures_util::StreamExt; use BuilderUpdate::*; // Wait for the build to finish or for it to emit a status message let update = tokio::select! { Some(progress) = self.rx.next() => progress, bundle = (&mut self.build_task) => { // Replace the build with an infinitely pending task so we can select it again without worrying about deadlocks/spins self.build_task = tokio::task::spawn(std::future::pending()); match bundle { Ok(Ok(bundle)) => BuilderUpdate::BuildReady { bundle }, Ok(Err(err)) => BuilderUpdate::BuildFailed { err }, Err(err) => BuilderUpdate::BuildFailed { err: crate::Error::Runtime(format!("Build panicked! {:#?}", err)) }, } }, Some(Ok(Some(msg))) = OptionFuture::from(self.stdout.as_mut().map(|f| f.next_line())) => { StdoutReceived { msg } }, Some(Ok(Some(msg))) = OptionFuture::from(self.stderr.as_mut().map(|f| f.next_line())) => { StderrReceived { msg } }, Some(status) = OptionFuture::from(self.child.as_mut().map(|f| f.wait())) => { // Panicking here is on purpose. If the task crashes due to a JoinError (a panic), // we want to propagate that panic up to the serve controller. let status = status.unwrap(); self.child = None; ProcessExited { status } } }; // Update the internal stage of the build so the UI can render it // *VERY IMPORTANT* - DO NOT AWAIT HERE // doing so will cause the changes to be lost since this wait call is called under a cancellable task // todo - move this handling to a separate function that won't be cancelled match &update { BuilderUpdate::Progress { stage } => { // Prevent updates from flowing in after the build has already finished if !self.is_finished() { self.stage = stage.clone(); match stage { BuildStage::Initializing => { self.compiled_crates = 0; self.bundling_progress = 0.0; } BuildStage::Starting { crate_count, .. } => { self.expected_crates = *crate_count; } BuildStage::InstallingTooling => {} BuildStage::Compiling { current, total, .. } => { self.compiled_crates = *current; self.expected_crates = *total; if self.compile_start.is_none() { self.compile_start = Some(Instant::now()); } } BuildStage::Bundling => { self.complete_compile(); self.bundling_progress = 0.0; self.bundle_start = Some(Instant::now()); } BuildStage::OptimizingWasm => {} BuildStage::CopyingAssets { current, total, .. } => { self.bundling_progress = *current as f64 / *total as f64; } BuildStage::Success => { self.compiled_crates = self.expected_crates; self.bundling_progress = 1.0; } BuildStage::Failed => { self.compiled_crates = self.expected_crates; self.bundling_progress = 1.0; } BuildStage::Aborted => {} BuildStage::Restarting => { self.compiled_crates = 0; self.expected_crates = 1; self.bundling_progress = 0.0; } BuildStage::RunningBindgen => {} _ => {} } } } BuilderUpdate::CompilerMessage { .. } => {} BuilderUpdate::BuildReady { .. } => { self.compiled_crates = self.expected_crates; self.bundling_progress = 1.0; self.stage = BuildStage::Success; self.complete_compile(); self.bundle_end = Some(Instant::now()); } BuilderUpdate::BuildFailed { .. } => { tracing::debug!("Setting builder to failed state"); self.stage = BuildStage::Failed; } StdoutReceived { .. } => {} StderrReceived { .. } => {} ProcessExited { .. } => {} } update } pub(crate) fn patch_rebuild(&mut self, changed_files: Vec) { // We need the rustc args from the original build to pass to the new build let Some(artifacts) = self.artifacts.as_ref().cloned() else { tracing::warn!("Ignoring patch rebuild since there is no existing build."); return; }; // On web, our patches are fully relocatable, so we don't need to worry about ASLR, but // for all other platforms, we need to use the ASLR reference to know where to insert the patch. let aslr_reference = match self.aslr_reference { Some(val) => val, None if self.build.platform == Platform::Web => 0, None => { tracing::warn!( "Ignoring hotpatch since there is no ASLR reference. Is the client connected?" ); return; } }; let cache = artifacts .patch_cache .clone() .context("Failed to get patch cache") .unwrap(); // Abort all the ongoing builds, cleaning up any loose artifacts and waiting to cleanly exit self.abort_all(BuildStage::Restarting); self.build_task = tokio::spawn({ let request = self.build.clone(); let ctx = BuildContext { tx: self.tx.clone(), mode: BuildMode::Thin { changed_files, rustc_args: artifacts.direct_rustc, aslr_reference, cache, }, }; async move { request.build(&ctx).await } }); } /// Restart this builder with new build arguments. pub(crate) fn start_rebuild(&mut self, mode: BuildMode) { // Abort all the ongoing builds, cleaning up any loose artifacts and waiting to cleanly exit // And then start a new build, resetting our progress/stage to the beginning and replacing the old tokio task self.abort_all(BuildStage::Restarting); self.artifacts.take(); self.patch_cache.take(); self.build_task = tokio::spawn({ let request = self.build.clone(); let ctx = BuildContext { tx: self.tx.clone(), mode, }; async move { request.build(&ctx).await } }); } /// Shutdown the current build process /// /// todo: might want to use a cancellation token here to allow cleaner shutdowns pub(crate) fn abort_all(&mut self, stage: BuildStage) { self.stage = stage; self.compiled_crates = 0; self.expected_crates = 1; self.bundling_progress = 0.0; self.compile_start = None; self.bundle_start = None; self.bundle_end = None; self.compile_end = None; self.build_task.abort(); } /// Wait for the build to finish, returning the final bundle /// Should only be used by code that's not interested in the intermediate updates and only cares about the final bundle /// /// todo(jon): maybe we want to do some logging here? The build/bundle/run screens could be made to /// use the TUI output for prettier outputs. pub(crate) async fn finish_build(&mut self) -> Result { loop { match self.wait().await { BuilderUpdate::Progress { stage } => { match &stage { BuildStage::Compiling { current, total, krate, .. } => { tracing::info!("Compiled [{current:>3}/{total}]: {krate}"); } BuildStage::RunningBindgen => tracing::info!("Running wasm-bindgen..."), BuildStage::CopyingAssets { current, total, path, } => { tracing::info!( "Copying asset ({}/{total}): {}", current + 1, path.display() ); } BuildStage::Bundling => tracing::info!("Bundling app..."), _ => {} } tracing::info!(json = ?StructuredOutput::BuildUpdate { stage: stage.clone() }); } BuilderUpdate::CompilerMessage { message } => { tracing::info!(json = ?StructuredOutput::CargoOutput { message: message.clone() }, %message); } BuilderUpdate::BuildReady { bundle } => { tracing::debug!(json = ?StructuredOutput::BuildFinished { path: self.build.root_dir(), }); return Ok(bundle); } BuilderUpdate::BuildFailed { err } => { // Flush remaining compiler messages while let Ok(Some(msg)) = self.rx.try_next() { if let BuilderUpdate::CompilerMessage { message } = msg { tracing::info!(json = ?StructuredOutput::CargoOutput { message: message.clone() }, %message); } } tracing::error!(?err, json = ?StructuredOutput::Error { message: err.to_string() }); return Err(err); } BuilderUpdate::StdoutReceived { .. } => {} BuilderUpdate::StderrReceived { .. } => {} BuilderUpdate::ProcessExited { .. } => {} } } } pub(crate) async fn open( &mut self, devserver_ip: SocketAddr, open_address: Option, start_fullstack_on_address: Option, open_browser: bool, always_on_top: bool, build_id: BuildId, ) -> Result<()> { let krate = &self.build; // Set the env vars that the clients will expect // These need to be stable within a release version (ie 0.6.0) let mut envs = vec![ (dioxus_cli_config::CLI_ENABLED_ENV, "true".to_string()), ( dioxus_cli_config::DEVSERVER_IP_ENV, devserver_ip.ip().to_string(), ), ( dioxus_cli_config::DEVSERVER_PORT_ENV, devserver_ip.port().to_string(), ), ( dioxus_cli_config::APP_TITLE_ENV, krate.config.web.app.title.clone(), ), ( dioxus_cli_config::SESSION_CACHE_DIR, self.build.session_cache_dir().display().to_string(), ), (dioxus_cli_config::BUILD_ID, build_id.0.to_string()), ( dioxus_cli_config::ALWAYS_ON_TOP_ENV, always_on_top.to_string(), ), // unset the cargo dirs in the event we're running `dx` locally // since the child process will inherit the env vars, we don't want to confuse the downstream process ("CARGO_MANIFEST_DIR", "".to_string()), ("RUST_BACKTRACE", "1".to_string()), ]; if let Some(base_path) = krate.base_path() { envs.push((dioxus_cli_config::ASSET_ROOT_ENV, base_path.to_string())); } if let Some(env_filter) = env::var_os("RUST_LOG").and_then(|e| e.into_string().ok()) { envs.push(("RUST_LOG", env_filter)); } // Launch the server if we were given an address to start it on, and the build includes a server. After we // start the server, consume its stdout/stderr. if let Some(addr) = start_fullstack_on_address { envs.push((dioxus_cli_config::SERVER_IP_ENV, addr.ip().to_string())); envs.push((dioxus_cli_config::SERVER_PORT_ENV, addr.port().to_string())); } // We try to use stdin/stdout to communicate with the app let running_process = match self.build.platform { // Unfortunately web won't let us get a proc handle to it (to read its stdout/stderr) so instead // use use the websocket to communicate with it. I wish we could merge the concepts here, // like say, opening the socket as a subprocess, but alas, it's simpler to do that somewhere else. Platform::Web => { // Only the first build we open the web app, after that the user knows it's running if open_browser { self.open_web(open_address.unwrap_or(devserver_ip)); } None } Platform::Ios => Some(self.open_ios_sim(envs).await?), Platform::Android => { self.open_android_sim(false, devserver_ip, envs).await?; None } // These are all just basically running the main exe, but with slightly different resource dir paths Platform::Server | Platform::MacOS | Platform::Windows | Platform::Linux | Platform::Liveview => Some(self.open_with_main_exe(envs)?), }; // If we have a running process, we need to attach to it and wait for its outputs if let Some(mut child) = running_process { let stdout = BufReader::new(child.stdout.take().unwrap()); let stderr = BufReader::new(child.stderr.take().unwrap()); self.stdout = Some(stdout.lines()); self.stderr = Some(stderr.lines()); self.child = Some(child); } self.builds_opened += 1; Ok(()) } /// Gracefully kill the process and all of its children /// /// Uses the `SIGTERM` signal on unix and `taskkill` on windows. /// This complex logic is necessary for things like window state preservation to work properly. /// /// Also wipes away the entropy executables if they exist. pub(crate) async fn soft_kill(&mut self) { use futures_util::FutureExt; // Kill any running executables on Windows let Some(mut process) = self.child.take() else { return; }; let Some(pid) = process.id() else { _ = process.kill().await; return; }; // on unix, we can send a signal to the process to shut down #[cfg(unix)] { _ = Command::new("kill") .args(["-s", "TERM", &pid.to_string()]) .spawn(); } // on windows, use the `taskkill` command #[cfg(windows)] { _ = Command::new("taskkill") .args(["/F", "/PID", &pid.to_string()]) .spawn(); } // join the wait with a 100ms timeout futures_util::select! { _ = process.wait().fuse() => {} _ = tokio::time::sleep(std::time::Duration::from_millis(1000)).fuse() => {} }; // Wipe out the entropy executables if they exist if let Some(entropy_app_exe) = self.entropy_app_exe.take() { _ = std::fs::remove_file(entropy_app_exe); } } pub(crate) async fn hotpatch( &mut self, res: &BuildArtifacts, cache: &HotpatchModuleCache, ) -> Result { let original = self.build.main_exe(); let new = self.build.patch_exe(res.time_start); let triple = self.build.triple.clone(); let original_artifacts = self.artifacts.as_ref().unwrap(); let asset_dir = self.build.asset_dir(); for (k, bundled) in res.assets.assets.iter() { let k = dunce::canonicalize(k)?; if original_artifacts.assets.assets.contains_key(k.as_path()) { continue; } let from = k.clone(); let to = asset_dir.join(bundled.bundled_path()); tracing::debug!("Copying asset from patch: {}", k.display()); if let Err(e) = dioxus_cli_opt::process_file_to(bundled.options(), &from, &to) { tracing::error!("Failed to copy asset: {e}"); continue; } // If the emulator is android, we need to copy the asset to the device with `adb push asset /data/local/tmp/dx/assets/filename.ext` if self.build.platform == Platform::Android { let changed_file = dunce::canonicalize(k).inspect_err(|e| { tracing::debug!("Failed to canonicalize hotreloaded asset: {e}") })?; let bundled_name = PathBuf::from(bundled.bundled_path()); _ = self .copy_file_to_android_tmp(&changed_file, &bundled_name) .await; } } tracing::debug!("Patching {} -> {}", original.display(), new.display()); let mut jump_table = crate::build::create_jump_table(&new, &triple, cache)?; // If it's android, we need to copy the assets to the device and then change the location of the patch if self.build.platform == Platform::Android { jump_table.lib = self .copy_file_to_android_tmp(&new, &(PathBuf::from(new.file_name().unwrap()))) .await?; } // Rebase the wasm binary to be relocatable once the jump table is generated if triple.architecture == target_lexicon::Architecture::Wasm32 { // Make sure we use the dir relative to the public dir, so the web can load it as a proper URL // // ie we would've shipped `/Users/foo/Projects/dioxus/target/dx/project/debug/web/public/wasm/lib.wasm` // but we want to ship `/wasm/lib.wasm` jump_table.lib = PathBuf::from("/").join(jump_table.lib.strip_prefix(self.build.root_dir()).unwrap()) } let changed_files = match &res.mode { BuildMode::Thin { changed_files, .. } => changed_files.clone(), _ => vec![], }; use crate::styles::{GLOW_STYLE, NOTE_STYLE}; let changed_file = changed_files.first().unwrap(); tracing::info!( "Hot-patching: {NOTE_STYLE}{}{NOTE_STYLE:#} took {GLOW_STYLE}{:?}ms{GLOW_STYLE:#}", changed_file .display() .to_string() .trim_start_matches(&self.build.crate_dir().display().to_string()), SystemTime::now() .duration_since(res.time_start) .unwrap() .as_millis() ); self.patches.push(jump_table.clone()); Ok(jump_table) } /// Hotreload an asset in the running app. /// /// This will modify the build dir in place! Be careful! We generally assume you want all bundles /// to reflect the latest changes, so we will modify the bundle. /// /// However, not all platforms work like this, so we might also need to update a separate asset /// dir that the system simulator might be providing. We know this is the case for ios simulators /// and haven't yet checked for android. /// /// This will return the bundled name of the asset such that we can send it to the clients letting /// them know what to reload. It's not super important that this is robust since most clients will /// kick all stylsheets without necessarily checking the name. pub(crate) async fn hotreload_bundled_asset(&self, changed_file: &PathBuf) -> Option { let artifacts = self.artifacts.as_ref()?; // Use the build dir if there's no runtime asset dir as the override. For the case of ios apps, // we won't actually be using the build dir. let asset_dir = match self.runtime_asset_dir.as_ref() { Some(dir) => dir.to_path_buf().join("assets/"), None => self.build.asset_dir(), }; // Canonicalize the path as Windows may use long-form paths "\\\\?\\C:\\". let changed_file = dunce::canonicalize(changed_file) .inspect_err(|e| tracing::debug!("Failed to canonicalize hotreloaded asset: {e}")) .ok()?; // The asset might've been renamed thanks to the manifest, let's attempt to reload that too let resource = artifacts.assets.assets.get(&changed_file)?; let output_path = asset_dir.join(resource.bundled_path()); tracing::debug!("Hotreloading asset {changed_file:?} in target {asset_dir:?}"); // Remove the old asset if it exists _ = std::fs::remove_file(&output_path); // And then process the asset with the options into the **old** asset location. If we recompiled, // the asset would be in a new location because the contents and hash have changed. Since we are // hotreloading, we need to use the old asset location it was originally written to. let options = *resource.options(); let res = process_file_to(&options, &changed_file, &output_path); let bundled_name = PathBuf::from(resource.bundled_path()); if let Err(e) = res { tracing::debug!("Failed to hotreload asset {e}"); } // If the emulator is android, we need to copy the asset to the device with `adb push asset /data/local/tmp/dx/assets/filename.ext` if self.build.platform == Platform::Android { _ = self .copy_file_to_android_tmp(&changed_file, &bundled_name) .await; } Some(bundled_name) } /// Copy this file to the tmp folder on the android device, returning the path to the copied file /// /// When we push patches (.so), the runtime will dlopen the file from the tmp folder by first copying /// it to shared memory. This is a workaround since not all android devices will be rooted and we /// can't drop the file into the `/data/data/com.org.app/lib/` directory. pub(crate) async fn copy_file_to_android_tmp( &self, changed_file: &Path, bundled_name: &Path, ) -> Result { let target = dioxus_cli_config::android_session_cache_dir().join(bundled_name); tracing::debug!("Pushing asset to device: {target:?}"); let res = Command::new(&self.build.workspace.android_tools()?.adb) .arg("push") .arg(changed_file) .arg(&target) .output() .await .context("Failed to push asset to device"); if let Err(e) = res { tracing::debug!("Failed to push asset to device: {e}"); } Ok(target) } /// Open the native app simply by running its main exe /// /// Eventually, for mac, we want to run the `.app` with `open` to fix issues with `dylib` paths, /// but for now, we just run the exe directly. Very few users should be caring about `dylib` search /// paths right now, but they will when we start to enable things like swift integration. /// /// Server/liveview/desktop are all basically the same, though fn open_with_main_exe(&mut self, envs: Vec<(&str, String)>) -> Result { let main_exe = self.app_exe(); tracing::debug!("Opening app with main exe: {main_exe:?}"); let child = Command::new(main_exe) .envs(envs) .stderr(Stdio::piped()) .stdout(Stdio::piped()) .kill_on_drop(true) .spawn()?; Ok(child) } /// Open the web app by opening the browser to the given address. /// Check if we need to use https or not, and if so, add the protocol. /// Go to the basepath if that's set too. fn open_web(&self, address: SocketAddr) { let base_path = self.build.config.web.app.base_path.clone(); let https = self.build.config.web.https.enabled.unwrap_or_default(); let protocol = if https { "https" } else { "http" }; let base_path = match base_path.as_deref() { Some(base_path) => format!("/{}", base_path.trim_matches('/')), None => "".to_owned(), }; _ = open::that_detached(format!("{protocol}://{address}{base_path}")); } /// Use `xcrun` to install the app to the simulator /// With simulators, we're free to basically do anything, so we don't need to do any fancy codesigning /// or entitlements, or anything like that. /// /// However, if there's no simulator running, this *might* fail. /// /// TODO(jon): we should probably check if there's a simulator running before trying to install, /// and open the simulator if we have to. async fn open_ios_sim(&mut self, envs: Vec<(&str, String)>) -> Result { tracing::debug!("Installing app to simulator {:?}", self.build.root_dir()); let res = Command::new("xcrun") .arg("simctl") .arg("install") .arg("booted") .arg(self.build.root_dir()) .output() .await?; tracing::debug!("Installed app to simulator with exit code: {res:?}"); // Remap the envs to the correct simctl env vars // iOS sim lets you pass env vars but they need to be in the format "SIMCTL_CHILD_XXX=XXX" let ios_envs = envs .iter() .map(|(k, v)| (format!("SIMCTL_CHILD_{k}"), v.clone())); let child = Command::new("xcrun") .arg("simctl") .arg("launch") .arg("--console") .arg("booted") .arg(self.build.bundle_identifier()) .envs(ios_envs) .stderr(Stdio::piped()) .stdout(Stdio::piped()) .kill_on_drop(true) .spawn()?; Ok(child) } /// We have this whole thing figured out, but we don't actually use it yet. /// /// Launching on devices is more complicated and requires us to codesign the app, which we don't /// currently do. /// /// Converting these commands shouldn't be too hard, but device support would imply we need /// better support for codesigning and entitlements. #[allow(unused)] async fn open_ios_device(&self) -> Result<()> { use serde_json::Value; let app_path = self.build.root_dir(); install_app(&app_path).await?; // 2. Determine which device the app was installed to let device_uuid = get_device_uuid().await?; // 3. Get the installation URL of the app let installation_url = get_installation_url(&device_uuid, &app_path).await?; // 4. Launch the app into the background, paused launch_app_paused(&device_uuid, &installation_url).await?; // 5. Pick up the paused app and resume it resume_app(&device_uuid).await?; async fn install_app(app_path: &PathBuf) -> Result<()> { let output = Command::new("xcrun") .args(["simctl", "install", "booted"]) .arg(app_path) .output() .await?; if !output.status.success() { return Err(format!("Failed to install app: {:?}", output).into()); } Ok(()) } async fn get_device_uuid() -> Result { let output = Command::new("xcrun") .args([ "devicectl", "list", "devices", "--json-output", "target/deviceid.json", ]) .output() .await?; let json: Value = serde_json::from_str(&std::fs::read_to_string("target/deviceid.json")?) .context("Failed to parse xcrun output")?; let device_uuid = json["result"]["devices"][0]["identifier"] .as_str() .ok_or("Failed to extract device UUID")? .to_string(); Ok(device_uuid) } async fn get_installation_url(device_uuid: &str, app_path: &Path) -> Result { // xcrun devicectl device install app --device --path --json-output let output = Command::new("xcrun") .args([ "devicectl", "device", "install", "app", "--device", device_uuid, &app_path.display().to_string(), "--json-output", "target/xcrun.json", ]) .output() .await?; if !output.status.success() { return Err(format!("Failed to install app: {:?}", output).into()); } let json: Value = serde_json::from_str(&std::fs::read_to_string("target/xcrun.json")?) .context("Failed to parse xcrun output")?; let installation_url = json["result"]["installedApplications"][0]["installationURL"] .as_str() .ok_or("Failed to extract installation URL")? .to_string(); Ok(installation_url) } async fn launch_app_paused(device_uuid: &str, installation_url: &str) -> Result<()> { let output = Command::new("xcrun") .args([ "devicectl", "device", "process", "launch", "--no-activate", "--verbose", "--device", device_uuid, installation_url, "--json-output", "target/launch.json", ]) .output() .await?; if !output.status.success() { return Err(format!("Failed to launch app: {:?}", output).into()); } Ok(()) } async fn resume_app(device_uuid: &str) -> Result<()> { let json: Value = serde_json::from_str(&std::fs::read_to_string("target/launch.json")?) .context("Failed to parse xcrun output")?; let status_pid = json["result"]["process"]["processIdentifier"] .as_u64() .ok_or("Failed to extract process identifier")?; let output = Command::new("xcrun") .args([ "devicectl", "device", "process", "resume", "--device", device_uuid, "--pid", &status_pid.to_string(), ]) .output() .await?; if !output.status.success() { return Err(format!("Failed to resume app: {:?}", output).into()); } Ok(()) } unimplemented!("dioxus-cli doesn't support ios devices yet.") } #[allow(unused)] async fn codesign_ios(&self) -> Result<()> { const CODESIGN_ERROR: &str = r#"This is likely because you haven't - Created a provisioning profile before - Accepted the Apple Developer Program License Agreement The agreement changes frequently and might need to be accepted again. To accept the agreement, go to https://developer.apple.com/account To create a provisioning profile, follow the instructions here: https://developer.apple.com/documentation/xcode/sharing-your-teams-signing-certificates"#; let profiles_folder = dirs::home_dir() .context("Your machine has no home-dir")? .join("Library/MobileDevice/Provisioning Profiles"); if !profiles_folder.exists() || profiles_folder.read_dir()?.next().is_none() { tracing::error!( r#"No provisioning profiles found when trying to codesign the app. We checked the folder: {} {CODESIGN_ERROR} "#, profiles_folder.display() ) } let identities = Command::new("security") .args(["find-identity", "-v", "-p", "codesigning"]) .output() .await .context("Failed to run `security find-identity -v -p codesigning`") .map(|e| { String::from_utf8(e.stdout) .context("Failed to parse `security find-identity -v -p codesigning`") })??; // Parsing this: // 1231231231231asdasdads123123 "Apple Development: foo@gmail.com (XYZYZY)" let app_dev_name = regex::Regex::new(r#""Apple Development: (.+)""#) .unwrap() .captures(&identities) .and_then(|caps| caps.get(1)) .map(|m| m.as_str()) .context( "Failed to find Apple Development in `security find-identity -v -p codesigning`", )?; // Acquire the provision file let provision_file = profiles_folder .read_dir()? .flatten() .find(|entry| { entry .file_name() .to_str() .map(|s| s.contains("mobileprovision")) .unwrap_or_default() }) .context("Failed to find a provisioning profile. \n\n{CODESIGN_ERROR}")?; // The .mobileprovision file has some random binary thrown into into, but it's still basically a plist // Let's use the plist markers to find the start and end of the plist fn cut_plist(bytes: &[u8], byte_match: &[u8]) -> Option { bytes .windows(byte_match.len()) .enumerate() .rev() .find(|(_, slice)| *slice == byte_match) .map(|(i, _)| i + byte_match.len()) } let bytes = std::fs::read(provision_file.path())?; let cut1 = cut_plist(&bytes, b""#.as_bytes()) .context("Failed to parse .mobileprovision file")?; let sub_bytes = &bytes[(cut1 - 6)..cut2]; let mbfile: ProvisioningProfile = plist::from_bytes(sub_bytes).context("Failed to parse .mobileprovision file")?; #[derive(serde::Deserialize, Debug)] struct ProvisioningProfile { #[serde(rename = "TeamIdentifier")] team_identifier: Vec, #[serde(rename = "ApplicationIdentifierPrefix")] application_identifier_prefix: Vec, #[serde(rename = "Entitlements")] entitlements: Entitlements, } #[derive(serde::Deserialize, Debug)] struct Entitlements { #[serde(rename = "application-identifier")] application_identifier: String, #[serde(rename = "keychain-access-groups")] keychain_access_groups: Vec, } let entielements_xml = format!( r#" application-identifier {APPLICATION_IDENTIFIER} keychain-access-groups {APP_ID_ACCESS_GROUP}.* get-task-allow com.apple.developer.team-identifier {TEAM_IDENTIFIER} "#, APPLICATION_IDENTIFIER = mbfile.entitlements.application_identifier, APP_ID_ACCESS_GROUP = mbfile.entitlements.keychain_access_groups[0], TEAM_IDENTIFIER = mbfile.team_identifier[0], ); // write to a temp file let temp_file = tempfile::NamedTempFile::new()?; std::fs::write(temp_file.path(), entielements_xml)?; // codesign the app let output = Command::new("codesign") .args([ "--force", "--entitlements", temp_file.path().to_str().unwrap(), "--sign", app_dev_name, ]) .arg(self.build.root_dir()) .output() .await .context("Failed to codesign the app")?; if !output.status.success() { let stderr = String::from_utf8(output.stderr).unwrap_or_default(); return Err(format!("Failed to codesign the app: {stderr}").into()); } Ok(()) } /// Launch the Android simulator and deploy the application. /// /// This function handles the process of starting the Android simulator, installing the APK, /// forwarding the development server port, and launching the application on the simulator. /// /// The following `adb` commands are executed: /// /// 1. **Enable Root Access**: /// - `adb root`: Enables root access on the Android simulator, allowing for advanced operations like pushing files to restricted directories. /// /// 2. **Port Forwarding**: /// - `adb reverse tcp: tcp:`: Forwards the development server port from the host /// machine to the Android simulator, enabling communication between the app and the dev server. /// /// 3. **APK Installation**: /// - `adb install -r `: Installs the APK onto the Android simulator. The `-r` flag /// ensures that any existing installation of the app is replaced. /// /// 4. **Environment Variables**: /// - Writes environment variables to a `.env` file in the session cache directory. /// - `adb push `: Pushes the `.env` file to the Android device /// to configure runtime environment variables for the app. /// /// 5. **App Launch**: /// - `adb shell am start -n /`: Launches the app on the Android /// simulator. The `` and `` are derived from the app's configuration. /// /// # Notes /// /// - This function is asynchronous and spawns a background task to handle the simulator setup and app launch. /// - The Android tools (`adb`) must be available in the system's PATH for this function to work. /// - If the app fails to launch, errors are logged for debugging purposes. /// /// # Resources: /// - https://developer.android.com/studio/run/emulator-commandline async fn open_android_sim( &self, root: bool, devserver_socket: SocketAddr, envs: Vec<(&'static str, String)>, ) -> Result<()> { let apk_path = self.build.debug_apk_path(); let session_cache = self.build.session_cache_dir(); let full_mobile_app_name = self.build.full_mobile_app_name(); let adb = self.build.workspace.android_tools()?.adb.clone(); // Start backgrounded since .open() is called while in the arm of the top-level match let _handle: JoinHandle> = tokio::task::spawn(async move { // call `adb root` so we can push patches to the device if root { if let Err(e) = Command::new(&adb).arg("root").output().await { tracing::error!("Failed to run `adb root`: {e}"); } } let port = devserver_socket.port(); if let Err(e) = Command::new(&adb) .arg("reverse") .arg(format!("tcp:{}", port)) .arg(format!("tcp:{}", port)) .output() .await { tracing::error!("failed to forward port {port}: {e}"); } // Wait for device to be ready let cmd = Command::new(&adb) .arg("wait-for-device") .arg("shell") .arg(r#"while [[ -z $(getprop sys.boot_completed) ]]; do sleep 1; done;"#) .output(); let cmd_future = cmd.fuse(); pin_mut!(cmd_future); tokio::select! { _ = &mut cmd_future => {} _ = tokio::time::sleep(Duration::from_millis(50)) => { tracing::info!("Waiting for android emulator to be ready..."); _ = cmd_future.await; } } // Install // adb install -r app-debug.apk let res = Command::new(&adb) .arg("install") .arg("-r") .arg(apk_path) .output() .await?; let std_err = String::from_utf8_lossy(&res.stderr); if !std_err.is_empty() { tracing::error!("Failed to install apk with `adb`: {std_err}"); } // Clear the session cache dir on the device Command::new(&adb) .arg("shell") .arg("rm") .arg("-rf") .arg(dioxus_cli_config::android_session_cache_dir()) .output() .await?; // Write the env vars to a .env file in our session cache let env_file = session_cache.join(".env"); _ = std::fs::write( &env_file, envs.iter() .map(|(key, value)| format!("{key}={value}")) .collect::>() .join("\n"), ); // Push the env file to the device Command::new(&adb) .arg("push") .arg(env_file) .arg(dioxus_cli_config::android_session_cache_dir().join(".env")) .output() .await?; // eventually, use the user's MainActivity, not our MainActivity // adb shell am start -n dev.dioxus.main/dev.dioxus.main.MainActivity let activity_name = format!("{}/dev.dioxus.main.MainActivity", full_mobile_app_name,); let res = Command::new(&adb) .arg("shell") .arg("am") .arg("start") .arg("-n") .arg(activity_name) .output() .await?; let std_err = String::from_utf8_lossy(res.stderr.trim_ascii()); if !std_err.is_empty() { tracing::error!("Failed to start app with `adb`: {std_err}"); } Ok(()) }); Ok(()) } fn make_entropy_path(exe: &PathBuf) -> PathBuf { let id = uuid::Uuid::new_v4(); let name = id.to_string(); let some_entropy = name.split('-').next().unwrap(); // Make a copy of the server exe with a new name let entropy_server_exe = exe.with_file_name(format!( "{}-{}", exe.file_name().unwrap().to_str().unwrap(), some_entropy )); std::fs::copy(exe, &entropy_server_exe).unwrap(); entropy_server_exe } fn app_exe(&mut self) -> PathBuf { let mut main_exe = self.build.main_exe(); // The requirement here is based on the platform, not necessarily our current architecture. let requires_entropy = match self.build.platform { // When running "bundled", we don't need entropy Platform::Web => false, Platform::MacOS => false, Platform::Ios => false, Platform::Android => false, // But on platforms that aren't running as "bundled", we do. Platform::Windows => true, Platform::Linux => true, Platform::Server => true, Platform::Liveview => true, }; if requires_entropy || crate::devcfg::should_force_entropy() { // If we already have an entropy app exe, return it - this is useful for re-opening the same app if let Some(existing_app_exe) = self.entropy_app_exe.clone() { return existing_app_exe; } let entropy_app_exe = Self::make_entropy_path(&main_exe); self.entropy_app_exe = Some(entropy_app_exe.clone()); main_exe = entropy_app_exe; } main_exe } fn complete_compile(&mut self) { if self.compile_end.is_none() { self.compiled_crates = self.expected_crates; self.compile_end = Some(Instant::now()); } } /// Get the total duration of the build, if all stages have completed pub(crate) fn total_build_time(&self) -> Option { Some(self.compile_duration()? + self.bundle_duration()?) } pub(crate) fn compile_duration(&self) -> Option { Some( self.compile_end .unwrap_or_else(Instant::now) .duration_since(self.compile_start?), ) } pub(crate) fn bundle_duration(&self) -> Option { Some( self.bundle_end .unwrap_or_else(Instant::now) .duration_since(self.bundle_start?), ) } /// Return a number between 0 and 1 representing the progress of the app build pub(crate) fn compile_progress(&self) -> f64 { self.compiled_crates as f64 / self.expected_crates as f64 } pub(crate) fn bundle_progress(&self) -> f64 { self.bundling_progress } pub(crate) fn is_finished(&self) -> bool { match self.stage { BuildStage::Success => true, BuildStage::Failed => true, BuildStage::Aborted => true, BuildStage::Restarting => false, _ => false, } } /// Check if the queued build is blocking hotreloads pub(crate) fn can_receive_hotreloads(&self) -> bool { matches!(&self.stage, BuildStage::Success | BuildStage::Failed) } }