Core dump on Solaris 10 allocating/deallocating memory
Hi,
We have a customer who reports our application causes a core dump on the Solaris 10 hosts. However, the core dump is happened on some hosts, but not on all.
I looked at several core dumps generated on differnt time, the stack
calls are different, but they all start from a new thread creation, and
end at libc and SGI STL.
Here is one example of stack calls:
#0 0xff1547b0 in getusa () from /lib/libc.so.1
#1 0x660f7a28 in ? ()
#2 0xff154fd8 in getsystemTZ () from /lib/libc.so.1
#3 0xff154f14 in _tzload () from /lib/libc.so.1
#4 0x3647e8 in _STLD::__malloc_alloc<0>::deallocate (__p=0xa1db70)
at ../../../../../third_party/stlport/stl/_alloc.h:114
#5 0x365930 in _STLD::allocator<char>::deallocate (this=0x96b124,
__p=0xa1db70 "000287890408", __n=13) at
../../../../../third_party/stlport/stl/_alloc.h:360
#6 0x36558c in _STLD::_String_base<char, _STLD::allocator><char>
>::_M_deallocate_block (this=0x96b11c) at
../../../../../third_party/stlport/stl/_string.h:140
#7 0x364f0c in _STLD::_String_base<char, _STLD::allocator><char>
>::~_String_base (this=0x96b11c, __in_chrg=2) at
../../../../../third_party/stlport/stl/_string.h:151
...
#16 0x39c374 in _STLD::__vector<device, _STLD::allocator><device>
>::clear (this=0xfed7f7e4)
at ../../../../../third_party/stlport/stl/_vector.h:488
#17 0x39bb44 in _STLD::vector<device, _STLD::allocator><device> >::clear
(this=0xfed7f7e4)
at ../../../../../third_party/stlport/stl/debug/_vector.h:262
...
#22 0x203590 in app_ThreadManager::_launch (pVoid=0x86d6c8) at
appthrdmgr.cxx:125
#23 0x240d40 in _internal_thread_wrapper (parameter=0x9c6bf8)
at appthrd.cxx:207
Other core dumps are generated when "new" or "free" are called. All the
cores are the same at the top ("#0" to "#3") and bottom (#22, #23).
Our application has no problem on other Solaris versions. Maybe our application has some problem in memory usage. But we could not reproduce this in our environment easily. Any suggests about what may be the problem?
Thanks.
