[mvapich-discuss] build error with --enable-thread-cs=per-object

[Jan David Mol]

Hi,

I'm experiencing build errors (both v1.9 release and svn trunk r6510) 
when enabling per-object locking (see information attached below).

It seems that the function MPIU_Thread_CS_enter_lockname_impl_ is used 
in mpiimpl.h before it is defined (in mpiimplthreadpost.h).

Cheers,
   Jan David Mol

In file included from ./src/include/mpiimpl.h:3821:0,
                  from ./src/mpid/ch3/include/mpidimpl.h:36,
                  from 
/src/mpid/ch3/channels/mrail/include/mpidi_ch3_impl.h:23,
                  from 
src/mpid/ch3/channels/common/src/affinity/hwloc_bind.c:13:
/src/include/mpiimplthreadpost.h:27:1: warning: conflicting types for 
'MPIU_Thread_CS_enter_lockname_impl_' [enabled by default]
/src/include/mpiimplthreadpost.h:27:1: error: static declaration of 
'MPIU_Thread_CS_enter_lockname_impl_' follows non-static declaration
/src/include/mpiimpl.h:1319:1: note: previous implicit declaration of 
'MPIU_Thread_CS_enter_lockname_impl_' was here
/src/include/mpiimplthreadpost.h:46:1: warning: conflicting types for 
'MPIU_Thread_CS_exit_lockname_impl_' [enabled by default]
/src/include/mpiimplthreadpost.h:46:1: error: static declaration of 
'MPIU_Thread_CS_exit_lockname_impl_' follows non-static declaration
/src/include/mpiimpl.h:1319:1: note: previous implicit declaration of 
'MPIU_Thread_CS_exit_lockname_impl_' was here

I'm building with:

/configure --prefix=/globalhome/mol/root-cobalt 
--with-cuda=/usr/local/cuda-5.5 
--with-cuda-include=/usr/local/cuda-5.5/targets/x86_64-linux/include/ 
--enable-shared --with-ib-libpath=/usr/lib64 --enable-threads=runtime 
--enable-handle-allocation=tls --with-hwloc 
--enable-thread-cs=per-object -enable-ref-count=lock-free --enable-cuda 
--with-atomic-primitives --enable-fast=all,O3 && make -j 32

Using

gcc version 4.6.3 (Ubuntu/Linaro 4.6.3-1ubuntu5)