DevOps, CI/CD and Automation

hi, i get a problem using OCI to connect to server. On X64 platform, Solaris10 system, and 64bit Oracle10.2 release, when i use OCIServerAttach() to attach a server, it will core dump. However, with the same code and same Oracle version, on a spark platform, it works well.My code is 32bit, so it depens on $ORACLE_HOME/lib32/libclntsh.so.10.1. Under Spark platform it works,but under x64 platform it will coredump. I copied a libclntsh.so.10.1 from a 32bit Oracle 10.2 client instance $ORACLE_HOME/lib on X86 platform to the X64 $ORACLE_HOME/lib32, the code then works. Why?
My code:

OCIEnv *m_pEnvHandle;
OCIError *m_pErrHandle;
OCIServer *m_pSrvHandle;
OCISvcCtx *m_pSvcHandle;

const char* i_pszSrv = "iccshis";

OCIEnvCreate(&m_pEnvHandle, OCI_THREADED|OCI_OBJECT, (dvoid *)0,
0, 0, 0, (size_t) 0, (dvoid **)0);
OCIHandleAlloc((dvoid*)m_pEnvHandle, (dvoid**)&m_pErrHandle,
OCI_HTYPE_ERROR, (size_t)0, (dvoid**)0);
OCIHandleAlloc((dvoid*)m_pEnvHandle, (dvoid**)&m_pSrvHandle,
OCI_HTYPE_SERVER, (size_t)0, (dvoid**)0);
OCIServerAttach(m_pSrvHandle, m_pErrHandle, (text*)i_pszSrv, strlen((char*)i_pszSrv), OCI_DEFAULT);

The stack back trace when the programme core dumped:
00000000 ???????? (fb1f3430, 0, 804068c, 822e1d8, b, 8040690)
fa996a3f nnfgrne (fb1f3430, 8043058, 0, fb1b5e38, 0, 8040e00) + 25b
faa0b30d nlolgobj (fb1f3430, 8043058, 16, 8040e64, 8040e68) + 669
fa9951ab nnfun2a (fb1f3430, 8043058, ff, 8043034, 8040ef0, 1000) + 15b
fa994e9e nnfsn2a (fb1f3430, 8043058, ff, 8043034, 8040ef0, 1000) + 36
fa983f97 niqname (fb1f3430, 1001, 8043058, 7, c0e887c, 8046290) + 2fb
fa8cfd38 kwfnran (9721ba8, 7, 804628c, 8046290, c0ee8bc, c0ee8c0) + 168
fa8a7b33 kwfcinit (c0e475c, 9721ba8, 7, 804628c, 8046290, be896d0) + 1b7
fa726f04 kpuatch (c0e475c, c0e41d0, 9721ba8, 7, 0, 0) + 7c0
fa7e6f06 OCIServerAttach (c0e475c, c0e41d0, 9721ba8, 7, 0, 0) + 56
f7574b07 bool COracleDBI::Open(const char*,const char*,const char*) (be7c998, 9721ba8, 89b6d90, 95f7170) + 12f;

Can anybody tell me why? thanks.