I wanted to make sure that a thread can not be cancelled asynchronous when it is in the cleanup push call, but i think it could be done better with InterlockedExchangePointer.
I will review and send an update of my pending patches with the current CVS sources tonight. Thomas On Mon, 10 Jun 2002, Robert Collins wrote: > Oh, and I'm not 100% sure that the cleanup_handler_push needs the locks > - shouldn't it be non cancellable? The IEEE P1003.1 reference I'm using > does not list pthread_clean_push as being cancellable, and explicitly > states that non listed functions (from the standard) are not > cancellable. > > Rob > > > -----Original Message----- > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED]] On Behalf Of Robert Collins > > Sent: Monday, 10 June 2002 12:24 PM > > To: 'Thomas Pfaff'; [EMAIL PROTECTED] > > Subject: RE: [PATCH] added locks in pthread code > > > > > > I'm applying a variation on this. Again, mainly OOP style changes, but > > also making the mutex an instance rather than pointer. (And where you > > aware that you where leaking the mutex?) > > > > Rob > > > > > -----Original Message----- > > > From: Thomas Pfaff [mailto:[EMAIL PROTECTED]] > > > Sent: Thursday, 25 April 2002 7:33 PM > > > To: [EMAIL PROTECTED] > > > Subject: [PATCH] added locks in pthread code > > > > > > > > > The patch will add locks via mutex around critical code to > > > protect against > > > race conditions and fix __pthread_detach to cleanup when thread has > > > already terminated. This an incremental update again. > > > > > > Greetings, > > > Thomas > > > > > > 2002-04-25 Thomas Pfaff <[EMAIL PROTECTED]> > > > > > > * thread.h (pthread::mutex): new member > > > * thread.cc (pthread::pthread): Set mutex to NULL. > > > (pthread::~pthread): Destroy mutex. > > > (pthread::create): Initialize mutex. > > > (thread_init_wrapper): Protect against race. > > > (__pthread_cleanup_push): Ditto. > > > (__pthread_exit): Ditto. > > > (__pthread_join): Ditto > > > (__pthread_detach): Protect against race and cleanup if > > > thread has > > > already terminated. > > > > > > > >
