[MPICH] Error compiling mpich2 on solaris with gcc, PTRACE_XXXX
James S Perrin
james.s.perrin at manchester.ac.uk
Thu Aug 2 09:14:37 CDT 2007
Hi,
OK thanks I was following up an early suggestion that it might work
if I used gcc.
Regards
James
Darius Buntinas wrote:
>
> The ssm channel is not supported on Solaris. We hope to have Solaris
> support for the Nemesis channel soon. Like ssm, Nemesis uses shared
> memory for intranode communication and networks for internode
> communication.
>
> Darius
>
> On 08/02/2007 08:58 AM, James S Perrin wrote:
>> Hi,
>>
>> I'm using :
>> mpich2-1.05p4 source
>> gcc 3.3.2
>> Solris 9 (sunOS 5.9)
>>
>> Configured as:
>> ./configure --prefix=/opt/apl/sp/mpich2 --with-mpe --with-pm=mpd
>> --with-device=ch3:ssm --with-thread-package=pthreads --disable-f77
>> --disable-f90 --disable-cxx
>>
>> make fails with the following error:
>>
>> gcc -DHAVE_CONFIG_H -I.
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/util/shm
>> -I../../../../include -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/include
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/include
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/include
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/datatype
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/datatype
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/channels/ssm/include
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/channels/ssm/include
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/locks
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/locks
>> -DUSE_PROCESS_LOCKS
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/sock
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/sock
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/sock/poll
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/common/sock/poll
>> -I/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/util/sock -c shmproc.c
>> shmproc.c: In function `MPIDI_SHM_AttachProc':
>> shmproc.c:59: error: `PTRACE_ATTACH' undeclared (first use in this
>> function)
>> shmproc.c:59: error: (Each undeclared identifier is reported only once
>> shmproc.c:59: error: for each function it appears in.)
>> shmproc.c: In function `MPIDI_SHM_DetachProc':
>> shmproc.c:80: error: `PTRACE_DETACH' undeclared (first use in this
>> function)
>> shmproc.c: In function `MPIDI_SHM_ReadProcessMemory':
>> shmproc.c:115: error: `PTRACE_PEEKDATA' undeclared (first use in this
>> function)
>> shmproc.c:115: warning: passing arg 3 of `ptrace' makes integer from
>> pointer without a cast
>> make[5]: *** [shmproc.o] Error 1
>> make[5]: Leaving directory
>> `/hpc_system/fnskgt/mpich2-1.0.5p4/src/mpid/ch3/util/shm'
>>
>>
>> The PTRACE_XXXX defines do not seem to be set on solaris, it uses
>> explicit integer values for ptrace requests according to the man page.
>>
>> shmproc.c has :
>>
>> #ifdef HAVE_SYS_PTRACE_H
>> #include <sys/ptrace.h>
>> #endif
>>
>> but does not #if..def the calls to ptrace or define PTRACE_XXXX
>> instead. Looking at the source for shmproc.c in CVS this still seems
>> to be the case (though it has moved to src/mpid/ch3/util/shmproc).
>>
>> From the little I understand after reading the man page only
>> PTRACE_PEEKDATA seems to have and equivalent value on this system - 2.
>> So there doesn't seem to be much that can be done, the man page in
>> courages use of /proc for debugging purposes instead. So this is
>> probably just a FYI.
>>
>> I've successfully compiled using SUN Studio but MPICH2 doesn't
>> currently support the ssm device for these compilers.
>>
>> Regards
>> James
>>
>
--
------------------------------------------------------------------------
James S. Perrin, | email: james.perrin at manchester.ac.uk
Manchester Visualization Centre, | web: www.mc.manchester.ac.uk
Kilburn Building, The University, | tel: +44 161 275 6945
Manchester, England. M13 9PL. | fax: +44 161 275 0637
------------------------------------------------------------------------
"The test of intellect is the refusal to belabour the obvious"
- Alfred Bester
------------------------------------------------------------------------
More information about the mpich-discuss
mailing list