Mailing-List: contact cygwin-help@sourceware.cygnus.com; run by ezmlm List-Subscribe: List-Archive: List-Post: List-Help: , Sender: cygwin-owner@sources.redhat.com Delivered-To: mailing list cygwin@sources.redhat.com Subject: pshared mutex's and cond variables - does anyone use them? From: Robert Collins To: Frederic Devernay Cc: cygwin@cygwin.com, Olivier Bantiche In-Reply-To: <20010830094600.A9594@perf.inria.fr> References: <20010830094600.A9594@perf.inria.fr> Content-Type: text/plain Content-Transfer-Encoding: 7bit X-Mailer: Evolution/0.12 (Preview Release) Date: 30 Aug 2001 19:40:19 +1000 Message-Id: <999164420.1059.3.camel@lifelesswks> Mime-Version: 1.0 On 30 Aug 2001 09:46:00 +0200, Frederic Devernay wrote: > Hello, > > there are officially two ways to initialize a pthread mutex: > pthread_mutex_t mut = PTHREAD_MUTEX_INITIALIZER; > or > pthread_mutex_t mut; > pthread_mutex_init(&mut,NULL); > > the problem is that the pthread_mutex_init in cygwin reads the of-course > uninitialized memory: > > as seen in cygwin-1.3.2-1/winsup/cygwin/thread.cc: > > __pthread_mutex_init (pthread_mutex_t * mutex, > const pthread_mutexattr_t * attr) > { > if ((((pshared_mutex *)(mutex))->flags & SYS_BASE == SYS_BASE)) > // a pshared mutex > return EBUSY; > ... > > If found this bug not by declaring a local variable as above, but by allocating > the mutex. the only workaround I found was to memset the memory area before > calling pthread_mutex_init. > > I don't know what a pshared_mutex is, but pthread_mutex_init should really > not assume that anything is valid in the given mutex! Yes, this is a known bug, last discussed right here about a week ago!. I'm seriously considering pulling the pshared code out of pthreads again. They cannot be implemented cleanly pending some significant work within cygwin1.dll, which has no timeframe (by this I mean a daemon for 'kernel-style' activity). They add significantly to the internal complexity in the meantime... Also note, that it is always possible to confuse pthread_mutex_init, on most platforms, but it shouldn't be possible to cause a segfault or anything nasty like that :]. Rob -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Bug reporting: http://cygwin.com/bugs.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/