Failure in class lookup

[Ravindra]

Hi,

 

We have an application developed using GNUStep(Solaris platform). While starting number of instances of this application at a time, we encountered two or three Servers that were not responsive.  In one particular test,  we started 80 Servers (starting 5 processes at a time with a delay of 60 Secs between next 5 processes so on). After all the processes were started, two or three Servers that were not responsive. When the test was repeated number of times, we observed that always 2 to 3 instances are not responsive, but the instances are not same.

We forced the Server to dump core by sending it a SIGBUS signal several times. The core always showed that numerous threads were stopped in calls with traces ending in:

 

#0  0xfea9b3e0 in _lwp_sema_wait ()

#1  0xfecc9820 in _park ()

#2  0xfecc94f8 in _swtch ()

#3  0xfeccade8 in _mutex_adaptive_lock ()

#4  0xfeccb7e8 in pthread_mutex_lock ()
#5  0xfee60504 in __objc_mutex_lock ()

#6  0xfee609dc in objc_mutex_lock ()

#7  0xfee59c00 in objc_get_class ()

 

A number of calls from our code that led up to this ending.  So, it seems like the underlying mechanism that GNUstep is using to serialize class lookups is somehow failing and a lock is getting locked and never getting unlocked at that level.

 

Any idea of how to resolve this problem.

 

Thanks,

Ravindra.