[omniNotify] Memory Leaks (both omniORB and omniNotify)

Frederic Prin frederic.prin at silvaco.com
Fri May 28 16:05:04 BST 2004


Hello,

While using omniNotify2.0beta with omniORB-4.0.3 on solaris 2.8, I
notice that notifd grows each time a push_structured_event is sent. I
use sdtprocess to track memory usage. I purify notifd and found some
PLK.
- Do you experience some memory leaks in notifd ?
- Do the latest 2.0 Release fix some memory leaks ?
- Do the latest omniORB snapshot fix some memory leaks ?

Thank you very much for your help.

Fred


This is the purify message (I now it's PLK not MLK but I seems to be
true)

         PLK: 11536 bytes potentially leaked in 412 blocks
         This memory was allocated from:
               malloc         [rtlib.o]
               c2n6Fi_Pv___1  [libCrun.so.1]
               void*operator new(unsigned) [rtlib.o]
               void*operator new[](unsigned) [rtlib.o]
 
IOP::TaggedProfile*_CORBA_Sequence<IOP::TaggedProfile>::allocbuf(unsigne
d long) [libomniORB4_4_0_3_R.a]
               void
_CORBA_Sequence<IOP::TaggedProfile>::copybuffer(unsigned long)
[libomniORB4_4_0_3_R.a]
         Block of 28 bytes (412 times); last block at 0x147c0d0

or

         PLK: 3592 bytes potentially leaked at 0xed7940
         This memory was allocated from:
               malloc         [rtlib.o]
               c2n6Fi_Pv___1  [libCrun.so.1]
               void*operator new(unsigned) [rtlib.o]
               void*operator new[](unsigned) [rtlib.o]
 
EventChannel_i::EventChannel_i(EventChannelFactory_i*,FilterFactory_i*,c
onst CosNotification::PropertySeq&,const
CosNotification::PropertySeq&,RDI_ServerQoS*,const long&)
[EventChannel_i.cc:123]
                  
                    // Initialize all statistics-related structures
                  
               =>   _thread_stats = new RDI_ThStat[RDI_TH_ARRAY_SZ];
                    RDI_AssertAllocThrowNo(_thread_stats, "Memory
allocation failed - RDI_ThStat object\n");
                    for (unsigned int id = 0; id < RDI_TH_ARRAY_SZ;
id++) {
                      _thread_stats[id]._num_rdi_match     = 0;
 
EventChannel_i*EventChannelFactory_i::_create_channel(long&)
[EventChannel_i.cc:1811]





More information about the omninotify-list mailing list