Files
react-native/ReactCommon/hermes/executor/HermesExecutorFactory.cpp
T
Eloy Durán 941bc0ec19 Upstream RN macOS Hermes integration bits (#29748)
Summary:
Microsoft’s RN for macOS fork supports the Hermes engine nowadays https://github.com/microsoft/react-native-macos/pull/473. As a longer term work item, we’ve started moving bits that are not invasive for iOS but _are_ a maintenance burden on us—mostly when merging—upstream. Seeing as this one is a recent addition, it seemed like a good candidate to start with.

As to the actual changes, these include:

* Sharing Android’s Hermes executor with the objc side of the codebase.
* Adding a CocoaPods subspec to build the Hermes inspector source and its dependencies (`Folly/Futures`, `libevent`).
* Adding the bits to the Xcode build phase script that creates the JS bundle for release builds to compile Hermes bytecode and source-maps…
* …coincidentally it turns out that the Xcode build phase script did _not_ by default output source-maps for iOS, which is now fixed too.

All of the Hermes bits are automatically enabled, on macOS, when providing the `hermes-engine-darwin` [npm package](https://www.npmjs.com/package/hermes-engine-darwin) and enabling the Hermes pods.

## Changelog

[General] [Added] - Upstream RN macOS Hermes integration bits

Pull Request resolved: https://github.com/facebook/react-native/pull/29748

Test Plan:
Building RNTester for iOS and Android still works as before.

To test the actual changes themselves, you’ll have to use the macOS target in RNTester in the macOS fork, or create a new application from `master`:

<img width="812" alt="Screenshot 2020-08-18 at 16 55 06" src="https://user-images.githubusercontent.com/2320/90547606-160f6480-e18c-11ea-9a98-edbbaa755800.png">

Reviewed By: TheSavior

Differential Revision: D23304618

Pulled By: fkgozali

fbshipit-source-id: 4ef0e0f60d909f3c59f9cfc87c667189df656a3b
2020-08-27 01:18:33 -07:00

236 lines
7.9 KiB
C++

/*
* Copyright (c) Facebook, Inc. and its affiliates.
*
* This source code is licensed under the MIT license found in the
* LICENSE file in the root directory of this source tree.
*/
#include "HermesExecutorFactory.h"
#include <thread>
#include <cxxreact/MessageQueueThread.h>
#include <cxxreact/SystraceSection.h>
#include <hermes/hermes.h>
#include <jsi/decorator.h>
#ifdef HERMES_ENABLE_DEBUGGER
#include <hermes/inspector/RuntimeAdapter.h>
#include <hermes/inspector/chrome/Registration.h>
#endif
#include "JSITracing.h"
using namespace facebook::hermes;
using namespace facebook::jsi;
namespace facebook {
namespace react {
namespace {
std::unique_ptr<HermesRuntime> makeHermesRuntimeSystraced(
const ::hermes::vm::RuntimeConfig &runtimeConfig) {
SystraceSection s("HermesExecutorFactory::makeHermesRuntimeSystraced");
return hermes::makeHermesRuntime(runtimeConfig);
}
#ifdef HERMES_ENABLE_DEBUGGER
class HermesExecutorRuntimeAdapter
: public facebook::hermes::inspector::RuntimeAdapter {
public:
HermesExecutorRuntimeAdapter(
std::shared_ptr<Runtime> runtime,
HermesRuntime &hermesRuntime,
std::shared_ptr<MessageQueueThread> thread)
: runtime_(runtime),
hermesRuntime_(hermesRuntime),
thread_(std::move(thread)) {}
virtual ~HermesExecutorRuntimeAdapter() = default;
jsi::Runtime &getRuntime() override {
return *runtime_;
}
debugger::Debugger &getDebugger() override {
return hermesRuntime_.getDebugger();
}
void tickleJs() override {
// The queue will ensure that runtime_ is still valid when this
// gets invoked.
thread_->runOnQueue([&runtime = runtime_]() {
auto func =
runtime->global().getPropertyAsFunction(*runtime, "__tickleJs");
func.call(*runtime);
});
}
private:
std::shared_ptr<Runtime> runtime_;
HermesRuntime &hermesRuntime_;
std::shared_ptr<MessageQueueThread> thread_;
};
#endif
struct ReentrancyCheck {
// This is effectively a very subtle and complex assert, so only
// include it in builds which would include asserts.
#ifndef NDEBUG
ReentrancyCheck() : tid(std::thread::id()), depth(0) {}
void before() {
std::thread::id this_id = std::this_thread::get_id();
std::thread::id expected = std::thread::id();
// A note on memory ordering: the main purpose of these checks is
// to observe a before/before race, without an intervening after.
// This will be detected by the compare_exchange_strong atomicity
// properties, regardless of memory order.
//
// For everything else, it is easiest to think of 'depth' as a
// proxy for any access made inside the VM. If access to depth
// are reordered incorrectly, the same could be true of any other
// operation made by the VM. In fact, using acquire/release
// memory ordering could create barriers which mask a programmer
// error. So, we use relaxed memory order, to avoid masking
// actual ordering errors. Although, in practice, ordering errors
// of this sort would be surprising, because the decorator would
// need to call after() without before().
if (tid.compare_exchange_strong(
expected, this_id, std::memory_order_relaxed)) {
// Returns true if tid and expected were the same. If they
// were, then the stored tid referred to no thread, and we
// atomically saved this thread's tid. Now increment depth.
assert(depth == 0 && "No thread id, but depth != 0");
++depth;
} else if (expected == this_id) {
// If the stored tid referred to a thread, expected was set to
// that value. If that value is this thread's tid, that's ok,
// just increment depth again.
assert(depth != 0 && "Thread id was set, but depth == 0");
++depth;
} else {
// The stored tid was some other thread. This indicates a bad
// programmer error, where VM methods were called on two
// different threads unsafely. Fail fast (and hard) so the
// crash can be analyzed.
__builtin_trap();
}
}
void after() {
assert(
tid.load(std::memory_order_relaxed) == std::this_thread::get_id() &&
"No thread id in after()");
if (--depth == 0) {
// If we decremented depth to zero, store no-thread into tid.
std::thread::id expected = std::this_thread::get_id();
bool didWrite = tid.compare_exchange_strong(
expected, std::thread::id(), std::memory_order_relaxed);
assert(didWrite && "Decremented to zero, but no tid write");
}
}
std::atomic<std::thread::id> tid;
// This is not atomic, as it is only written or read from the owning
// thread.
unsigned int depth;
#endif
};
// This adds ReentrancyCheck and debugger enable/teardown to the given
// Runtime.
class DecoratedRuntime : public jsi::WithRuntimeDecorator<ReentrancyCheck> {
public:
// The first argument may be another decorater which itself
// decorates the real HermesRuntime, depending on the build config.
// The second argument is the real HermesRuntime as well to
// manage the debugger registration.
DecoratedRuntime(
std::unique_ptr<Runtime> runtime,
HermesRuntime &hermesRuntime,
std::shared_ptr<MessageQueueThread> jsQueue)
: jsi::WithRuntimeDecorator<ReentrancyCheck>(*runtime, reentrancyCheck_),
runtime_(std::move(runtime)),
hermesRuntime_(hermesRuntime) {
#ifdef HERMES_ENABLE_DEBUGGER
auto adapter = std::make_unique<HermesExecutorRuntimeAdapter>(
runtime_, hermesRuntime_, jsQueue);
facebook::hermes::inspector::chrome::enableDebugging(
std::move(adapter), "Hermes React Native");
#else
(void)hermesRuntime_;
#endif
}
~DecoratedRuntime() {
#ifdef HERMES_ENABLE_DEBUGGER
facebook::hermes::inspector::chrome::disableDebugging(hermesRuntime_);
#endif
}
private:
// runtime_ is a potentially decorated Runtime.
// hermesRuntime is a reference to a HermesRuntime managed by runtime_.
//
// HermesExecutorRuntimeAdapter requirements are kept, because the
// dtor will disable debugging on the HermesRuntime before the
// member managing it is destroyed.
std::shared_ptr<Runtime> runtime_;
ReentrancyCheck reentrancyCheck_;
HermesRuntime &hermesRuntime_;
};
} // namespace
std::unique_ptr<JSExecutor> HermesExecutorFactory::createJSExecutor(
std::shared_ptr<ExecutorDelegate> delegate,
std::shared_ptr<MessageQueueThread> jsQueue) {
std::unique_ptr<HermesRuntime> hermesRuntime =
makeHermesRuntimeSystraced(runtimeConfig_);
HermesRuntime &hermesRuntimeRef = *hermesRuntime;
auto decoratedRuntime = std::make_shared<DecoratedRuntime>(
std::move(hermesRuntime), hermesRuntimeRef, jsQueue);
// So what do we have now?
// DecoratedRuntime -> HermesRuntime
//
// DecoratedRuntime is held by JSIExecutor. When it gets used, it
// will check that it's on the right thread, do any necessary trace
// logging, then call the real HermesRuntime. When it is destroyed,
// it will shut down the debugger before the HermesRuntime is. In
// the normal case where debugging is not compiled in,
// all that's left is the thread checking.
// Add js engine information to Error.prototype so in error reporting we
// can send this information.
auto errorPrototype =
decoratedRuntime->global()
.getPropertyAsObject(*decoratedRuntime, "Error")
.getPropertyAsObject(*decoratedRuntime, "prototype");
errorPrototype.setProperty(*decoratedRuntime, "jsEngine", "hermes");
return std::make_unique<HermesExecutor>(
decoratedRuntime, delegate, jsQueue, timeoutInvoker_, runtimeInstaller_);
}
HermesExecutor::HermesExecutor(
std::shared_ptr<jsi::Runtime> runtime,
std::shared_ptr<ExecutorDelegate> delegate,
std::shared_ptr<MessageQueueThread> jsQueue,
const JSIScopedTimeoutInvoker &timeoutInvoker,
RuntimeInstaller runtimeInstaller)
: JSIExecutor(runtime, delegate, timeoutInvoker, runtimeInstaller) {
jsi::addNativeTracingHooks(*runtime);
}
} // namespace react
} // namespace facebook