// #TODO: #OS_Win32
// [ ] #Thread cleanup: in `thread_deinit` is there any requirement to cleanup child threads?
// [ ] #Exception handling code in `Win32_Exception_Filter`
// [ ] #cpuid - enumerate CPUs and Thread count (current implementation doesn't work)

#if OS_WINDOWS
constexpr s64 FILETIME_TO_UNIX = 116444736000000000i64;
f64 GetUnixTimestamp () {
    FILETIME fileTime;
    GetSystemTimePreciseAsFileTime(&fileTime);
    s64 ticks = ((s64)fileTime.dwHighDateTime << (s64)32) | (s64)fileTime.dwLowDateTime;
    return (ticks - FILETIME_TO_UNIX) / (10.0 * 1000.0 * 1000.0);
}
s64 GetUnixTimestampNanoseconds () {
    FILETIME fileTime;
    GetSystemTimePreciseAsFileTime(&fileTime);
    
    s64 ticks = ((s64)fileTime.dwHighDateTime << (s64)32) 
                | (s64)fileTime.dwLowDateTime;  // in 100ns ticks
    s64 unix_time = (ticks - FILETIME_TO_UNIX); // in 100ns ticks
    
    s64 unix_time_nanoseconds = unix_time * 100;
    
    return unix_time_nanoseconds;
}
#endif

struct OS_System_Info {
  s32 logical_processor_count;
  s32 physical_core_count;
  s32 primary_core_count;
  s32 secondary_core_count; // Any weaker or "Efficiency" cores.
  u64 page_size;
  u64 large_page_size;
  u64 allocation_granularity;
  string machine_name;
};

struct OS_Process_Info {
  u32 process_id;
  b32 large_pages_allowed;
  string binary_path;
  string working_path;
  string user_program_data_path;
  Array<string> module_load_paths;
  Array<string> environment_paths;
};

struct OS_State_Win32 {
  Arena* arena;
  
  OS_System_Info system_info;
  OS_Process_Info process_info;
};

global OS_State_Win32 os_state_w32;
internal b32 win32_g_is_quiet = 0; // No console output

internal LONG WINAPI Win32_Exception_Filter (EXCEPTION_POINTERS* exception_ptrs) {
  if (win32_g_is_quiet) { ExitProcess(1); }

  local_persist volatile LONG first = 0;
  if(InterlockedCompareExchange(&first, 1, 0) != 0)
  { // prevent failures in other threads to popup same message box
    // this handler just shows first thread that crashes
    // we are terminating afterwards anyway
    for (;;) Sleep(1000);
  }
  
  // #Exception handling code (TODO)
  
  return 0;
}

// internal void Main_Entry_Point (int argc, WCHAR **argv);
internal void Win32_Entry_Point (int argc, WCHAR **argv) {
  // See: w32_entry_point_caller(); (raddebugger)
  SetUnhandledExceptionFilter(&Win32_Exception_Filter);
  
  SYSTEM_INFO sysinfo = {0};
  GetSystemInfo(&sysinfo);
  
  // Try to allow large pages if we can.
  // b32 large_pages_allowed = 0;
  // {
  //   HANDLE token;
  //   if(OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &token))
  //   {
  //     LUID luid;
  //     if(LookupPrivilegeValue(0, SE_LOCK_MEMORY_NAME, &luid))
  //     {
  //       TOKEN_PRIVILEGES priv;
  //       priv.PrivilegeCount           = 1;
  //       priv.Privileges[0].Luid       = luid;
  //       priv.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED;
  //       large_pages_allowed = !!AdjustTokenPrivileges(token, 0, &priv, sizeof(priv), 0, 0);
  //     }
  //     CloseHandle(token);
  //   }
  // }
  
  Bootstrap_Main_Thread_Context();
  
  push_arena(get_thread_context()->arena);
  
  { OS_System_Info* info = &os_state_w32.system_info;
    info->logical_processor_count = (s32)sysinfo.dwNumberOfProcessors;
    info->page_size               = sysinfo.dwPageSize;
    info->large_page_size         = GetLargePageMinimum();
    info->allocation_granularity  = sysinfo.dwAllocationGranularity;
  }
  { OS_Process_Info* info = &os_state_w32.process_info;
    info->large_pages_allowed = false;
    info->process_id = GetCurrentProcessId();
  }
  // #cpuid
  /*{ OS_System_Info* info = &os_state_w32.system_info;
    // [ ] Extract input args 
    u32 length;
    GetLogicalProcessorInformationEx(RelationProcessorCore, nullptr, (PDWORD)&length);
    u8* cpu_information_buffer = NewArray<u8>(length);
    GetLogicalProcessorInformationEx(RelationProcessorCore, // *sigh* 
      (PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX)cpu_information_buffer, (PDWORD)&length);
    
    u32 offset = 0;
    
    u32 all_cpus_count = 0;
    u32 max_performance = 0;
    u32 performance_core_count = 0;
    // u32 efficient_core_count;
    
    while (offset < length) {
      SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX* cpu_information 
        = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)(cpu_information_buffer + offset);
      
      offset += cpu_information->Size;
      
      u32 count_per_group_physical = 1;
      u32 value = (u32)cpu_information->Processor.GroupMask->Mask;
      u32 count_per_group = __popcnt(value); // logical
      if (cpu_information->Relationship != RelationProcessorCore) continue;
      
      if (cpu_information->Processor.EfficiencyClass > max_performance) {
        max_performance = cpu_information->Processor.EfficiencyClass;
        performance_core_count = count_per_group_physical;
      } else if (cpu_information->Processor.EfficiencyClass == max_performance) {
        performance_core_count += count_per_group_physical;
      }
      
      all_cpus_count += count_per_group;  
    }
  
    info->physical_core_count = (s32)all_cpus_count;
    info->primary_core_count  = (s32)performance_core_count;
  } 
  // info->secondary_core_count = ;
  */
  { OS_System_Info* info = &os_state_w32.system_info;
    u8 buffer[MAX_COMPUTERNAME_LENGTH + 1] = {0};
    DWORD size = MAX_COMPUTERNAME_LENGTH + 1;
    if(GetComputerNameA((char*)buffer, &size)) {
      string machine_name_temp = string(size, buffer);
      info->machine_name = copy_string(machine_name_temp);
    }
  }
  
  { OS_Process_Info* info = &os_state_w32.process_info;
    DWORD length = GetCurrentDirectoryW(0, 0);
    // This can be freed later when we call temp_reset();
    u16* memory = NewArray<u16>(get_temp_allocator(), length + 1);
    length = GetCurrentDirectoryW(length + 1, (WCHAR*)memory);
    info->working_path = wide_to_utf8(memory, length);
    Assert(is_valid(info->working_path));
  }
  
  // [ ] Get Working directory (info->working_path)
  // [ ] GetEnvironmentStringsW
  // temp_reset();
  printf("Hello there!\n\n");
}

C_LINKAGE DWORD OS_Windows_Thread_Entry_Point (void* parameter) {
  Thread* thread = (Thread*)parameter;

  set_thread_context(thread->context);
  
  DWORD result = (DWORD)thread->proc(thread);
  
  return result;
}

// Individual Thread API
internal bool thread_init (Thread* thread, Thread_Proc proc, string thread_name="") {
  Assert(thread != nullptr && proc != nullptr);
  
  DWORD windows_thread_id = 0;

  HANDLE windows_thread = CreateThread(nullptr, 0, OS_Windows_Thread_Entry_Point, 
    thread, CREATE_SUSPENDED, &windows_thread_id);
  
  if (windows_thread == 0 || windows_thread == INVALID_HANDLE_VALUE) {
    return false;
  }
  
  s64 this_thread_index = InterlockedIncrement(&next_thread_index);
  
  ExpandableArena* arena_ex = expandable_arena_new(Arena_Reserve::Size_64M, 16);
  
  thread->context = New<Thread_Context>(get_allocator(arena_ex));
  thread->context->temp        = expandable_arena_new(Arena_Reserve::Size_2M, 16);
  thread->context->arena       = arena_ex;
  thread->context->allocator   = get_allocator(arena_ex);
  thread->context->thread_idx  = (s32)this_thread_index;
  push_arena(arena_ex);
  thread->context->thread_name = copy_string(thread_name);
  
  thread->os_thread.windows_thread    = windows_thread;
  thread->os_thread.windows_thread_id = windows_thread_id;
  
  thread->proc = proc;
  thread->index = this_thread_index;
  
  return true;
}

internal void thread_deinit (Thread* thread) {
  if (thread->os_thread.windows_thread) {
    CloseHandle(thread->os_thread.windows_thread);
  }
  thread->os_thread.windows_thread = nullptr;
  
  arena_delete(thread->context->temp);
  arena_delete(thread->context->arena);
}

internal void thread_start (Thread* thread) {
  ResumeThread(thread->os_thread.windows_thread);
}

internal bool thread_is_done (Thread* thread, s32 milliseconds=0) {
  Assert(milliseconds >= -1);
  
  DWORD result = WaitForSingleObject(thread->os_thread.windows_thread, (DWORD)milliseconds);
  return result != WAIT_TIMEOUT;
}

// #thread_primitives
internal void mutex_init (Mutex* mutex) {
  InitializeCriticalSection(&mutex->csection);
}
internal void mutex_destroy (Mutex* mutex) {
  DeleteCriticalSection(&mutex->csection);
}
internal void lock (Mutex* mutex) {
  EnterCriticalSection(&mutex->csection);
}
internal void unlock (Mutex* mutex) {
  LeaveCriticalSection(&mutex->csection);
}
internal void semaphore_init (Semaphore* sem, s32 initial_value) {
  Assert(initial_value >= 0);
  sem->event = CreateSemaphoreW(nullptr, initial_value, 0x7fffffff, nullptr);
}
internal void semaphore_destroy (Semaphore* sem) {
  CloseHandle(sem->event);
}
internal void signal (Semaphore* sem) {
  ReleaseSemaphore(sem->event, 1, nullptr);
}

internal Wait_For_Result wait_for (Semaphore* sem, s32 milliseconds) {
  DWORD res = 0;
  if (milliseconds < 0) {
    res = WaitForSingleObject(sem->event, INFINITE);
  } else {
    res = WaitForSingleObject(sem->event, (u32)milliseconds);
  }
  
  Assert(res != WAIT_FAILED);
  
  if (res == WAIT_OBJECT_0) return Wait_For_Result::SUCCESS;
  if (res == WAIT_TIMEOUT)  return Wait_For_Result::TIMEOUT;
  
  return Wait_For_Result::ERROR;
}

internal void condition_variable_init (Condition_Variable* cv) {
  InitializeConditionVariable(&cv->condition_variable);
}
internal void condition_variable_destroy (Condition_Variable* cv) {
  // No action required.
}
internal void wait (Condition_Variable* cv, Mutex* mutex, s32 wait_time_ms) {
  SleepConditionVariableCS(&cv->condition_variable, &mutex->csection, (DWORD)wait_time_ms);
}
internal void wake (Condition_Variable* cv) {
  WakeConditionVariable(&cv->condition_variable);
}
internal void wake_all (Condition_Variable* cv) {
  WakeAllConditionVariable(&cv->condition_variable);
}

// #window_creation
Window_Type create_window (string new_window_name) {
  return 0;
}

// #TODO: #window_creation
// [ ] resize_window
// [ ] position_window
// [ ] toggle_fullscreen
// [ ] get_dimensions

// #TODO: #window_interaction (mouse/keyboard)
// [ ] get_mouse_pointer_position
// [ ] ... What APIs do I need for Keyboard