Huaidong Qiu wrote:
\lib\dtrans\source\win32\dnd\target.cxx
void SAL_CALL DropTarget::initialize( const Sequence< Any >& aArguments )
throw(Exception, RuntimeException)
{
...........
m_hOleThread= CreateThread( NULL, 0,
(LPTHREAD_START_ROUTINE)DndTargetOleSTAFunc,
&m_evtThreadReady, 0,
&m_threadIdTarget);
WaitForSingleObject( m_evtThreadReady, INFINITE);
...........
}
DWORD WINAPI DndTargetOleSTAFunc(LPVOID pParams)
{
HRESULT hr= OleInitialize( NULL);
if( SUCCEEDED( hr) )
{
MSG msg;
// force the creation of a message queue
PeekMessage( &msg, (HWND)NULL, 0, 0, PM_NOREMOVE);
// Signal the creator ( DropTarget::initialize) that the thread is
// ready to receive messages.
SetEvent( *(HANDLE*) pParams);
...........
}
// no SetEvent( *(HANDLE*) pParams); action this branch.
}
If OleInitialize() inside DndTargetOleSTAFunc failed, DropTarget::initialize
will wait the m_evtThreadReady event forever. Could this cause freezing
problem?
Hi Huaidong,
I must confess that I am not very familiar with the DnD code. If
DropTarget::initialize is called from the main thread (UI thread) and
OleInitialize will fail then you are right that this would result in a
freeze. I checked several hundreds of freeze in OpenOffice.org annd I
have never seen your scenario. So I would think that theoretically your
assumption is true but why should OleInitialize fail in this special
scenario? The thread function calls OleInitialize the first time and
doesn't change anything. Did you see a freeze in the debugger where you
could these methods on the stack on two different thread?
Regards,
Carsten
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@framework.openoffice.org
For additional commands, e-mail: dev-h...@framework.openoffice.org
