fix bugs for waiter of linux (still has some bugs in win, multi-wait is TBD)

src/platform/waiter_linux.h

@@ -23,123 +23,161 @@
 namespace ipc {
 namespace detail {
 
-class semaphore {
+class mutex {
+    pthread_mutex_t mutex_ = PTHREAD_MUTEX_INITIALIZER;
+
 public:
-    using handle_t = sem_t*;
+    pthread_mutex_t& native() {
-
+        return mutex_;
-    constexpr static handle_t invalid() {
-        return SEM_FAILED;
     }
 
-    static handle_t open(char const* name) {
+    bool open() {
-        handle_t sem = ::sem_open(name, O_CREAT, 0666, 0);
+        int eno;
-        if (sem == SEM_FAILED) {
+        // init mutex
-            ipc::error("fail sem_open[%d]: %s\n", errno, name);
+        pthread_mutexattr_t mutex_attr;
-            return invalid();
+        if ((eno = ::pthread_mutexattr_init(&mutex_attr)) != 0) {
+            ipc::error("fail pthread_mutexattr_init[%d]\n", eno);
+            return false;
         }
-        return sem;
+        IPC_UNUSED_ auto guard_mutex_attr = unique_ptr(&mutex_attr, ::pthread_mutexattr_destroy);
+        if ((eno = ::pthread_mutexattr_setpshared(&mutex_attr, PTHREAD_PROCESS_SHARED)) != 0) {
+            ipc::error("fail pthread_mutexattr_setpshared[%d]\n", eno);
+            return false;
+        }
+        if ((eno = ::pthread_mutex_init(&mutex_, &mutex_attr)) != 0) {
+            ipc::error("fail pthread_mutex_init[%d]\n", eno);
+            return false;
+        }
+        return true;
     }
 
-#pragma push_macro("IPC_SEMAPHORE_FUNC_")
+    bool close() {
-#undef  IPC_SEMAPHORE_FUNC_
+        int eno;
-#define IPC_SEMAPHORE_FUNC_(CALL, PAR)              \
+        if ((eno = ::pthread_mutex_destroy(&mutex_)) != 0) {
-    if (::CALL(PAR) != 0) {                         \
+            ipc::error("fail pthread_mutex_destroy[%d]\n", eno);
-        ipc::error("fail " #CALL "[%d]\n", errno);  \
+            return false;
-        return false;                               \
+        }
-    }                                               \
+        return true;
-    return true
-
-    static bool close(handle_t h) {
-        if (h == invalid()) return false;
-        IPC_SEMAPHORE_FUNC_(sem_close, h);
     }
 
-    static bool destroy(char const* name) {
+    bool lock() {
-        IPC_SEMAPHORE_FUNC_(sem_unlink, name);
+        int eno;
+        if ((eno = ::pthread_mutex_lock(&mutex_)) != 0) {
+            ipc::error("fail pthread_mutex_lock[%d]\n", eno);
+            return false;
+        }
+        return true;
     }
 
-    static bool post(handle_t h) {
+    bool unlock() {
-        if (h == invalid()) return false;
+        int eno;
-        IPC_SEMAPHORE_FUNC_(sem_post, h);
+        if ((eno = ::pthread_mutex_unlock(&mutex_)) != 0) {
+            ipc::error("fail pthread_mutex_unlock[%d]\n", eno);
+            return false;
         }
-
+        return true;
-    static bool wait(handle_t h) {
-        if (h == invalid()) return false;
-        IPC_SEMAPHORE_FUNC_(sem_wait, h);
     }
-
-#pragma pop_macro("IPC_SEMAPHORE_FUNC_")
 };
 
-class event {
+class condition {
-    std::atomic<std::size_t>* cnt_ = nullptr;
+    pthread_cond_t cond_  = PTHREAD_COND_INITIALIZER;
-    semaphore::handle_t sem_ = semaphore::invalid();
-    std::size_t wait_id_;
-
-    std::string name() const {
-        return "__IPC_WAIT__" + std::to_string(wait_id_);
-    }
 
 public:
-    event(std::size_t id)
+    bool open() {
-        : wait_id_(static_cast<uint_t<16>>(id)) {
+        int eno;
-        auto n = name();
+        // init condition
-        cnt_ = static_cast<std::atomic<std::size_t>*>(
+        pthread_condattr_t cond_attr;
-                    shm::acquire(n.c_str(), sizeof(std::atomic<std::size_t>)));
+        if ((eno = ::pthread_condattr_init(&cond_attr)) != 0) {
-        if (cnt_ == nullptr) {
+            ipc::error("fail pthread_condattr_init[%d]\n", eno);
-            ipc::error("fail shm::acquire: %s\n", n.c_str());
+            return false;
-            return;
         }
-        cnt_->fetch_add(1, std::memory_order_acquire);
+        IPC_UNUSED_ auto guard_cond_attr = unique_ptr(&cond_attr, ::pthread_condattr_destroy);
-        sem_ = semaphore::open(n.c_str());
+        if ((eno = ::pthread_condattr_setpshared(&cond_attr, PTHREAD_PROCESS_SHARED)) != 0) {
+            ipc::error("fail pthread_condattr_setpshared[%d]\n", eno);
+            return false;
+        }
+        if ((eno = ::pthread_cond_init(&cond_, &cond_attr)) != 0) {
+            ipc::error("fail pthread_cond_init[%d]\n", eno);
+            return false;
+        }
+        return true;
     }
 
-    ~event() {
+    bool close() {
-        semaphore::close(sem_);
+        int eno;
-        if (cnt_->fetch_sub(1, std::memory_order_release) == 1) {
+        if ((eno = ::pthread_cond_destroy(&cond_)) != 0) {
-            semaphore::destroy(name().c_str());
+            ipc::error("fail pthread_cond_destroy[%d]\n", eno);
+            return false;
         }
-        shm::release(cnt_, sizeof(std::atomic<std::size_t>));
+        return true;
     }
 
-    auto get_id() const noexcept {
+    bool wait(mutex& mtx) {
-        return wait_id_;
+        int eno;
+        if ((eno = ::pthread_cond_wait(&cond_, &mtx.native())) != 0) {
+            ipc::error("fail pthread_cond_wait[%d]\n", eno);
+            return false;
         }
-
+        return true;
-    bool wait() {
-        if (sem_ == semaphore::invalid()) return false;
-        return semaphore::wait(sem_);
     }
 
     bool notify() {
-        if (sem_ == semaphore::invalid()) return false;
+        int eno;
-        return semaphore::post(sem_);
+        if ((eno = ::pthread_cond_signal(&cond_)) != 0) {
+            ipc::error("fail pthread_cond_signal[%d]\n", eno);
+            return false;
+        }
+        return true;
+    }
+
+    bool broadcast() {
+        int eno;
+        if ((eno = ::pthread_cond_broadcast(&cond_)) != 0) {
+            ipc::error("fail pthread_cond_broadcast[%d]\n", eno);
+            return false;
+        }
+        return true;
+    }
+};
+
+class semaphore {
+    mutex     lock_;
+    condition cond_;
+    long      counter_ = 0;
+
+public:
+    bool open() {
+        return lock_.open() && cond_.open();
+    }
+
+    void close() {
+        cond_.close();
+        lock_.close();
+    }
+
+    template <typename F>
+    void wait_if(F&& check) {
+        IPC_UNUSED_ auto guard = ipc::detail::unique_lock(lock_);
+        while ((counter_ <= 0) && std::forward<F>(check)()) {
+            cond_.wait(lock_);
+        }
+        -- counter_;
+    }
+
+    template <typename F>
+    bool post(F&& count) {
+        IPC_UNUSED_ auto guard = ipc::detail::unique_lock(lock_);
+        auto c = std::forward<F>(count)();
+        if (c <= 0) return false;
+        counter_ += c;
+        return cond_.broadcast();
     }
 };
 
 class waiter {
-    using evt_id_t = decltype(std::declval<event>().get_id());
+    semaphore sem_;
-
+    std::atomic<long>     counter_ { 0 };
-    std::atomic<unsigned> counter_ { 0 };
+    std::atomic<unsigned> opened_  { 0 };
-    spin_lock evt_lc_;
-    id_pool<sizeof(evt_id_t)> evt_ids_;
-
-    std::size_t push_event(event const & evt) {
-        IPC_UNUSED_ auto guard = ipc::detail::unique_lock(evt_lc_);
-        std::size_t id = evt_ids_.acquire();
-        if (id == invalid_value) {
-            ipc::error("fail push_event[has too many waiters]\n");
-            return id;
-        }
-        (*static_cast<evt_id_t*>(evt_ids_.at(id))) = evt.get_id();
-        evt_ids_.mark_acquired(id);
-        return id;
-    }
-
-    void pop_event(std::size_t id) {
-        IPC_UNUSED_ auto guard = ipc::detail::unique_lock(evt_lc_);
-        evt_ids_.release(id);
-    }
 
 public:
     using handle_t = waiter * ;
@@ -152,60 +190,48 @@ public:
         if (name == nullptr || name[0] == '\0') {
             return invalid();
         }
-        if (counter_.fetch_add(1, std::memory_order_acq_rel) == 0) {
+        if ((opened_.fetch_add(1, std::memory_order_acq_rel) == 0) && !sem_.open()) {
-            evt_ids_.init();
+            return invalid();
         }
         return this;
     }
 
     void close(handle_t h) {
         if (h == invalid()) return;
-        counter_.fetch_sub(1, std::memory_order_acq_rel);
+        if (opened_.fetch_sub(1, std::memory_order_release) == 1) {
+            sem_.close();
+        }
     }
 
-    static bool wait_all(std::tuple<waiter*, handle_t> const * all, std::size_t size) {
+    template <typename F>
+    static bool multi_wait_if(std::tuple<waiter*, handle_t> const * all, std::size_t size, F&& /*check*/) {
         if (all == nullptr || size == 0) {
             return false;
         }
-        // calc a new wait-id & construct event object
+        return true;
-        event evt { ipc::detail::calc_unique_id() };
-        auto ids = static_cast<std::size_t*>(mem::alloc(sizeof(std::size_t) * size));
-        for (std::size_t i = 0; i < size; ++i) {
-            ids[i] = std::get<0>(all[i])->push_event(evt);
-        }
-        IPC_UNUSED_ auto guard = unique_ptr(ids, [all, size](std::size_t* ids) {
-            for (std::size_t i = 0; i < size; ++i) {
-                std::get<0>(all[i])->pop_event(ids[i]);
-            }
-            mem::free(ids, sizeof(std::size_t) * size);
-        });
-        // wait for event signal
-        return evt.wait();
     }
 
-    bool wait(handle_t h) {
+    template <typename F>
+    bool wait_if(handle_t h, F&& check) {
         if (h == invalid()) return false;
-        auto info = std::make_tuple(this, h);
+        counter_.fetch_add(1, std::memory_order_release);
-        return wait_all(&info, 1);
+        IPC_UNUSED_ auto guard = unique_ptr(&counter_, [](decltype(counter_)* c) {
+            c->fetch_sub(1, std::memory_order_release);
+        });
+        sem_.wait_if(std::forward<F>(check));
+        return true;
     }
 
     void notify(handle_t h) {
         if (h == invalid()) return;
-        IPC_UNUSED_ auto guard = ipc::detail::unique_lock(evt_lc_);
+        sem_.post([this] {
-        evt_ids_.for_acquired([this](auto id) {
+            return (0 < counter_.load(std::memory_order_relaxed)) ? 1 : 0;
-            event evt { *static_cast<evt_id_t*>(evt_ids_.at(id)) };
-            return !evt.notify(); // return if succ
         });
     }
 
     void broadcast(handle_t h) {
         if (h == invalid()) return;
-        IPC_UNUSED_ auto guard = ipc::detail::unique_lock(evt_lc_);
+        sem_.post([this] { return counter_.load(std::memory_order_relaxed); });
-        evt_ids_.for_acquired([this](auto id) {
-            event evt { *static_cast<evt_id_t*>(evt_ids_.at(id)) };
-            evt.notify();
-            return true; // return after all
-        });
     }
 };
 

src/platform/waiter_win.h

@@ -37,10 +37,12 @@ public:
         ::CloseHandle(h);
     }
 
-    static bool wait_all(std::tuple<waiter*, handle_t> const * all, std::size_t size) {
+    template <typename F>
+    static bool multi_wait_if(std::tuple<waiter*, handle_t> const * all, std::size_t size, F&& check) {
         if (all == nullptr || size == 0) {
             return false;
         }
+        if (!std::forward<F>(check)()) return true;
         auto hs = static_cast<handle_t*>(mem::alloc(sizeof(handle_t) * size));
         IPC_UNUSED_ auto guard = unique_ptr(hs, [size](void* p) { mem::free(p, sizeof(handle_t) * size); });
         std::size_t i = 0;
@@ -55,8 +57,10 @@ public:
         return ::WaitForMultipleObjects(static_cast<DWORD>(i), hs, FALSE, INFINITE) != WAIT_FAILED;
     }
 
-    bool wait(handle_t h) {
+    template <typename F>
+    bool wait_if(handle_t h, F&& check) {
         if (h == invalid()) return false;
+        if (!std::forward<F>(check)()) return true;
         counter_.fetch_add(1, std::memory_order_relaxed);
         std::atomic_thread_fence(std::memory_order_release);
         return ::WaitForSingleObject(h, INFINITE) == WAIT_OBJECT_0;

src/platform/waiter_wrapper.h

@@ -42,6 +42,7 @@ public:
     waiter_t const * waiter() const { return w_; }
 
     void attach(waiter_t* w) {
+        close();
         w_ = w;
     }
 
@@ -62,7 +63,9 @@ public:
         h_ = waiter_t::invalid();
     }
 
-    bool wait_all(waiter_wrapper * all, std::size_t size) {
+
+    template <typename F>
+    bool multi_wait_if(waiter_wrapper * all, std::size_t size, F&& check) {
         if (all == nullptr || size == 0) {
             return false;
         }
@@ -75,12 +78,13 @@ public:
             if (!w.valid()) continue;
             hs[i] = w.to_w_info();
         }
-        return waiter_t::wait_all(hs, i);
+        return waiter_t::multi_wait_if(hs, i, std::forward<F>(check));
     }
 
-    bool wait() {
+    template <typename F>
+    bool wait_if(F&& check) {
         if (!valid()) return false;
-        return w_->wait(h_);
+        return w_->wait_if(h_, std::forward<F>(check));
     }
 
     bool notify() {

src/queue.h

@@ -110,7 +110,11 @@ public:
     bool wait_for_connect(Elems* elems, std::size_t count) {
         if (elems == nullptr) return false;
         for (unsigned k = 0; elems->conn_count() < count;) {
-            ipc::sleep(k, [this] { return cc_waiter_.wait(); });
+            ipc::sleep(k, [this, elems, count] {
+                return cc_waiter_.wait_if([elems, count] {
+                    return elems->conn_count() < count;
+                });
+            });
         }
         return true;
     }
@@ -214,13 +218,20 @@ public:
             return {};
         }
         T item;
-        for (unsigned k = 0;;) {
+        auto pop_item = [this, &item] {
-            if (elems_->pop(&this->cursor_, [&item](void* p) {
+            return elems_->pop(&this->cursor_, [&item](void* p) {
                 ::new (&item) T(std::move(*static_cast<T*>(p)));
-            })) {
+            });
-                return item;
+        };
-            }
+        for (unsigned k = 0;;) {
-            ipc::sleep(k, [this] { return this->waiter_.wait(); });
+            if (pop_item()) return item;
+            bool succ = false;
+            ipc::sleep(k, [this, &succ, &pop_item] {
+                return this->waiter_.wait_if([&succ, &pop_item] {
+                    return !(succ = pop_item());
+                });
+            });
+            if (succ) return item;
         }
     }
 };

src/waiter.cpp

@@ -64,7 +64,7 @@ void waiter::close() {
 }
 
 bool waiter::wait() {
-    return impl(p_)->w_.wait();
+    return impl(p_)->w_.wait_if([] { return true; });
 }
 
 bool waiter::notify() {

test/test_waiter.cpp

@@ -1,4 +1,5 @@
 #include <thread>
+#include <iostream>
 
 #include "platform/waiter_wrapper.h"
 #include "test.h"
@@ -13,34 +14,68 @@ class Unit : public TestSuite {
     }
 
 private slots:
-    void test_wakeup();
+    void test_broadcast();
+    void test_multiwait();
 } unit__;
 
 #include "test_waiter.moc"
 
-void Unit::test_wakeup() {
+void Unit::test_broadcast() {
     ipc::detail::waiter w;
+    std::thread ts[10];
 
-    std::thread t1 {[&w] {
+    for (auto& t : ts) {
+        t = std::thread([&w] {
             ipc::detail::waiter_wrapper wp { &w };
             QVERIFY(wp.open("test-ipc-waiter"));
-        QVERIFY(wp.wait());
+            QVERIFY(wp.wait_if([] { return true; }));
-    }};
+        });
-
+    }
-    std::thread t2 {[&w] {
-        ipc::detail::waiter_wrapper wp { &w };
-        QVERIFY(wp.open("test-ipc-waiter"));
-        QVERIFY(wp.wait());
-    }};
 
     ipc::detail::waiter_wrapper wp { &w };
     QVERIFY(wp.open("test-ipc-waiter"));
 
+    std::cout << "waiting for broadcast...\n";
     std::this_thread::sleep_for(std::chrono::seconds(1));
     QVERIFY(wp.broadcast());
 
-    t1.join();
+    for (auto& t : ts) t.join();
-    t2.join();
+}
+
+void Unit::test_multiwait() {
+    ipc::detail::waiter w, mw;
+    std::thread ts[10];
+    ipc::detail::waiter_wrapper ws[10];
+
+    std::size_t i = 0;
+    for (auto& t : ts) {
+        t = std::thread([&w, &mw, &ws, i] {
+            ipc::detail::waiter_wrapper wp { &w };
+            ws[i].attach(&mw);
+
+            QVERIFY(wp.open("test-ipc-waiter"));
+            QVERIFY(ws[i].open("test-ipc-multiwait"));
+
+            QVERIFY(wp.wait_if([] { return true; }));
+            if (i == 3) {
+                std::cout << "waiting for notify...\n";
+                std::this_thread::sleep_for(std::chrono::seconds(1));
+                QVERIFY(ws[i].broadcast());
+            }
+        });
+        ++i;
+    }
+
+    ipc::detail::waiter_wrapper wp { &w }, wm { &mw };
+    QVERIFY(wp.open("test-ipc-waiter"));
+    QVERIFY(wm.open("test-ipc-multiwait"));
+
+    std::cout << "waiting for broadcast...\n";
+    std::this_thread::sleep_for(std::chrono::seconds(1));
+    QVERIFY(wp.broadcast());
+    QVERIFY(wm.multi_wait_if(ws, 10, [] { return true; }));
+
+    for (auto& t : ts) t.join();
 }
 
 } // internal-linkage