Merge branch 'main' into to-globe-rectangle

CHANGES.md

@@ -7,6 +7,7 @@
 - Added `Uri::getPath` and `Uri::setPath`.
 - Added `TileTransform::setTransform`.
 - Added `BoundingRegionBuilder::toGlobeRectangle`.
+- Added `waitInMainThread` method to `Future` and `SharedFuture`.
 
 ##### Fixes :wrench:
 

CesiumAsync/include/CesiumAsync/Future.h

@@ -233,11 +233,32 @@ template <typename T> class Future final {
    * deadlock because the main thread tasks will never complete while this
    * method is blocking the main thread.
    *
+   * To wait in the main thread, use {@link waitInMainThread} instead.
+   *
    * @return The value if the future resolves successfully.
    * @throws An exception if the future rejected.
    */
   T wait() { return this->_task.get(); }
 
+  /**
+   * @brief Waits for this future to resolve or reject in the main thread while
+   * also processing main-thread tasks.
+   *
+   * This method must be called from the main thread.
+   *
+   * The function does not return until {@link Future::isReady} returns true.
+   * In the meantime, main-thread tasks are processed as necessary. This method
+   * does not spin wait; it suspends the calling thread by waiting on a
+   * condition variable when there is no work to do.
+   *
+   * @return The value if the future resolves successfully.
+   * @throws An exception if the future rejected.
+   */
+  T waitInMainThread() {
+    return this->_pSchedulers->mainThread.dispatchUntilTaskCompletes(
+        std::move(this->_task));
+  }
+
   /**
    * @brief Determines if this future is already resolved or rejected.
    *

CesiumAsync/include/CesiumAsync/Impl/QueuedScheduler.h

@@ -3,19 +3,88 @@
 #include "ImmediateScheduler.h"
 #include "cesium-async++.h"
 
+#include <atomic>
+
 namespace CesiumAsync {
 namespace CesiumImpl {
 
 class QueuedScheduler {
 public:
+  QueuedScheduler();
+  ~QueuedScheduler();
+
   void schedule(async::task_run_handle t);
   void dispatchQueuedContinuations();
   bool dispatchZeroOrOneContinuation();
 
+  template <typename T> T dispatchUntilTaskCompletes(async::task<T>&& task) {
+    // Set up a continuation to unblock the blocking dispatch when this task
+    // completes.
+    //
+    // We use the `isDone` flag as the loop termination condition to
+    // avoid a race condition that can lead to a deadlock. If we used
+    // `unblockTask.ready()` as the termination condition instead, then it's
+    // possible for events to happen as follows:
+    //
+    // 1. The original `task` completes in a worker thread and the `unblockTask`
+    // continuation is invoked immediately in the same thread.
+    // 2. The unblockTask continuation calls `unblock`, which terminates the
+    // `wait` on the condition variable in the main thread.
+    // 3. The main thread resumes and the while loop in this function spins back
+    // around and evaluates `unblockTask.ready()`. This returns false because
+    // the unblockTask continuation has not actually finished running in the
+    // worker thread yet. The main thread starts waiting on the condition
+    // variable again.
+    // 4. The `unblockTask` continuation finally finishes, making
+    // `unblockTask.ready()` return true, but it's too late. The main thread is
+    // already waiting on the condition variable.
+    //
+    // By setting the atomic `isDone` flag before calling `unblock`, we ensure
+    // that the loop termination condition is satisfied before the main thread
+    // is awoken, avoiding the potential deadlock.
+
+    std::atomic<bool> isDone = false;
+    async::task<T> unblockTask =
+        task.then(async::inline_scheduler(), [this, &isDone](auto&& value) {
+          isDone = true;
+          this->unblock();
+          return std::move(value);
+        });
+
+    while (!isDone) {
+      this->dispatchInternal(true);
+    }
+
+    return std::move(unblockTask).get();
+  }
+
+  template <typename T>
+  T dispatchUntilTaskCompletes(const async::shared_task<T>& task) {
+    // Set up a continuation to unblock the blocking dispatch when this task
+    // completes. This case is simpler than the one above because a SharedFuture
+    // supports multiple continuations. We can use readiness of the _original_
+    // task to terminate the loop while unblocking in a separate continuation
+    // guaranteed to run only after that termination condition is satisfied.
+    async::task<void> unblockTask =
+        task.then(async::inline_scheduler(), [this](const auto&) {
+          this->unblock();
+        });
+
+    while (!task.ready()) {
+      this->dispatchInternal(true);
+    }
+
+    return task.get();
+  }
+
   ImmediateScheduler<QueuedScheduler> immediate{this};
 
 private:
-  async::fifo_scheduler _scheduler;
+  bool dispatchInternal(bool blockIfNoTasks);
+  void unblock();
+
+  struct Impl;
+  std::unique_ptr<Impl> _pImpl;
 };
 
 } // namespace CesiumImpl

CesiumAsync/include/CesiumAsync/SharedFuture.h

@@ -217,6 +217,8 @@ template <typename T> class SharedFuture final {
    * deadlock because the main thread tasks will never complete while this
    * method is blocking the main thread.
    *
+   * To wait in the main thread, use {@link waitInMainThread} instead.
+   *
    * @return The value if the future resolves successfully.
    * @throws An exception if the future rejected.
    */
@@ -236,6 +238,25 @@ template <typename T> class SharedFuture final {
     this->_task.get();
   }
 
+  /**
+   * @brief Waits for this future to resolve or reject in the main thread while
+   * also processing main-thread tasks.
+   *
+   * This method must be called from the main thread.
+   *
+   * The function does not return until {@link Future::isReady} returns true.
+   * In the meantime, main-thread tasks are processed as necessary. This method
+   * does not spin wait; it suspends the calling thread by waiting on a
+   * condition variable when there is no work to do.
+   *
+   * @return The value if the future resolves successfully.
+   * @throws An exception if the future rejected.
+   */
+  T waitInMainThread() {
+    return this->_pSchedulers->mainThread.dispatchUntilTaskCompletes(
+        std::move(this->_task));
+  }
+
   /**
    * @brief Determines if this future is already resolved or rejected.
    *

CesiumAsync/src/AsyncSystem.cpp

@@ -2,8 +2,6 @@
 
 #include "CesiumAsync/ITaskProcessor.h"
 
-#include <future>
-
 namespace CesiumAsync {
 AsyncSystem::AsyncSystem(
     const std::shared_ptr<ITaskProcessor>& pTaskProcessor) noexcept

CesiumAsync/src/QueuedScheduler.cpp

@@ -1,17 +1,66 @@
 #include "CesiumAsync/Impl/QueuedScheduler.h"
 
-using namespace CesiumAsync::CesiumImpl;
+#include <condition_variable>
+#include <mutex>
+
+// Hackily use Async++'s internal fifo_queue. We could copy it instead - it's
+// not much code - but why create the duplication? However, we are assuming that
+// Async++'s source (not just headers) are available while building
+// cesium-native. If that's a problem in some context, we'll need to do that
+// duplication of fifo_queue after all.
+#include <async++/../../src/fifo_queue.h>
+
+namespace CesiumAsync::CesiumImpl {
+
+struct QueuedScheduler::Impl {
+  async::detail::fifo_queue queue;
+  std::mutex mutex;
+  std::condition_variable conditionVariable;
+};
+
+QueuedScheduler::QueuedScheduler() : _pImpl(std::make_unique<Impl>()) {}
+QueuedScheduler::~QueuedScheduler() = default;
 
 void QueuedScheduler::schedule(async::task_run_handle t) {
-  this->_scheduler.schedule(std::move(t));
+  std::unique_lock<std::mutex> guard(this->_pImpl->mutex);
+  this->_pImpl->queue.push(std::move(t));
+
+  // Notify listeners that there is new work.
+  this->_pImpl->conditionVariable.notify_all();
 }
 
 void QueuedScheduler::dispatchQueuedContinuations() {
-  auto scope = this->immediate.scope();
-  this->_scheduler.run_all_tasks();
+  while (this->dispatchZeroOrOneContinuation()) {
+  }
 }
 
 bool QueuedScheduler::dispatchZeroOrOneContinuation() {
-  auto scope = this->immediate.scope();
-  return this->_scheduler.try_run_one_task();
+  return this->dispatchInternal(false);
+}
+
+bool QueuedScheduler::dispatchInternal(bool blockIfNoTasks) {
+  async::task_run_handle t;
+
+  {
+    std::unique_lock<std::mutex> guard(this->_pImpl->mutex);
+    t = this->_pImpl->queue.pop();
+    if (blockIfNoTasks && !t) {
+      this->_pImpl->conditionVariable.wait(guard);
+    }
+  }
+
+  if (t) {
+    auto scope = this->immediate.scope();
+    t.run();
+    return true;
+  } else {
+    return false;
+  }
 }
+
+void QueuedScheduler::unblock() {
+  std::unique_lock<std::mutex> guard(this->_pImpl->mutex);
+  this->_pImpl->conditionVariable.notify_all();
+}
+
+} // namespace CesiumAsync::CesiumImpl

CesiumAsync/test/TestAsyncSystem.cpp

@@ -572,4 +572,96 @@ TEST_CASE("AsyncSystem") {
 
     CHECK(checksCompleted);
   }
+
+  SECTION("waitInMainThread") {
+    SECTION("Future returning a value") {
+      bool called = false;
+      Future<int> future =
+          asyncSystem.createResolvedFuture().thenInMainThread([&called]() {
+            called = true;
+            return 4;
+          });
+      int value = std::move(future).waitInMainThread();
+      CHECK(called);
+      CHECK(value == 4);
+    }
+
+    SECTION("Future returning void") {
+      bool called = false;
+      Future<void> future = asyncSystem.createResolvedFuture().thenInMainThread(
+          [&called]() { called = true; });
+      std::move(future).waitInMainThread();
+      CHECK(called);
+    }
+
+    SECTION("SharedFuture returning a value") {
+      bool called = false;
+      SharedFuture<int> future = asyncSystem.createResolvedFuture()
+                                     .thenInMainThread([&called]() {
+                                       called = true;
+                                       return 4;
+                                     })
+                                     .share();
+      int value = future.waitInMainThread();
+      CHECK(called);
+      CHECK(value == 4);
+    }
+
+    SECTION("SharedFuture returning void") {
+      bool called = false;
+      SharedFuture<void> future =
+          asyncSystem.createResolvedFuture()
+              .thenInMainThread([&called]() { called = true; })
+              .share();
+      future.waitInMainThread();
+      CHECK(called);
+    }
+
+    SECTION("Future resolving while main thread is waiting") {
+      bool called1 = false;
+      bool called2 = false;
+      Future<void> future =
+          asyncSystem.createResolvedFuture()
+              .thenInWorkerThread([&called1]() {
+                using namespace std::chrono_literals;
+                // should be long enough for the main thread to start waiting on
+                // the conditional, without slowing the test down too much.
+                std::this_thread::sleep_for(20ms);
+                called1 = true;
+              })
+              .thenInMainThread([&called2]() { called2 = true; });
+      future.waitInMainThread();
+      CHECK(called1);
+      CHECK(called2);
+    }
+
+    SECTION("Future resolving from a worker while main thread is waiting") {
+      bool called1 = false;
+      bool called2 = false;
+      bool called3 = false;
+      Future<void> future =
+          asyncSystem.createResolvedFuture()
+              .thenInWorkerThread([&called1]() {
+                using namespace std::chrono_literals;
+                // should be long enough for the main thread to start waiting on
+                // the conditional, without slowing the test down too much.
+                std::this_thread::sleep_for(20ms);
+                called1 = true;
+              })
+              .thenInMainThread([&called2]() { called2 = true; })
+              .thenInWorkerThread([&called3]() {
+                using namespace std::chrono_literals;
+                // Sufficient time for the main thread to drop back into waiting
+                // on the conditional again after it was awakened by the
+                // scheduling of the main thread continuation above. It should
+                // awaken again when this continuation completes.
+                std::this_thread::sleep_for(20ms);
+                called3 = true;
+              });
+      future.waitInMainThread();
+      CHECK(called1);
+      CHECK(called2);
+      CHECK(called3);
+    }
+  }
 }

CesiumGltf/generated/include/CesiumGltf/ExtensionCesiumPrimitiveOutline.h

@@ -0,0 +1,29 @@
+// This file was generated by generate-classes.
+// DO NOT EDIT THIS FILE!
+#pragma once
+
+#include "CesiumGltf/Library.h"
+
+#include <CesiumUtility/ExtensibleObject.h>
+
+#include <cstdint>
+
+namespace CesiumGltf {
+/**
+ * @brief glTF extension for indicating that some edges of a primitive's
+ * triangles should be outlined.
+ */
+struct CESIUMGLTF_API ExtensionCesiumPrimitiveOutline final
+    : public CesiumUtility::ExtensibleObject {
+  static inline constexpr const char* TypeName =
+      "ExtensionCesiumPrimitiveOutline";
+  static inline constexpr const char* ExtensionName =
+      "CESIUM_primitive_outline";
+
+  /**
+   * @brief The index of the accessor providing the list of highlighted lines at
+   * the edge of this primitive's triangles.
+   */
+  int32_t indices = -1;
+};
+} // namespace CesiumGltf

CesiumGltf/generated/include/CesiumGltf/ExtensionKhrDracoMeshCompression.h

@@ -11,7 +11,7 @@
 
 namespace CesiumGltf {
 /**
- * @brief KHR_draco_mesh_compression extension
+ * @brief KHR_draco_mesh_compression glTF Mesh Primitive Extension
  */
 struct CESIUMGLTF_API ExtensionKhrDracoMeshCompression final
     : public CesiumUtility::ExtensibleObject {

CesiumGltf/generated/include/CesiumGltf/ExtensionMeshPrimitiveKhrMaterialsVariants.h

@@ -11,7 +11,7 @@
 
 namespace CesiumGltf {
 /**
- * @brief KHR_materials_variants mesh primitive extension
+ * @brief KHR_materials_variants glTF Mesh Primitive Extension
  */
 struct CESIUMGLTF_API ExtensionMeshPrimitiveKhrMaterialsVariants final
     : public CesiumUtility::ExtensibleObject {