Better yet -- I never explicitly asked you what version you were using.
Shame on me. Looking in your config.log, it's 7.1.1.
This is a known problem in 7.1.1 -- the fix is in 7.1.2 (i.e., to
automatically decide that BSD's should use the "none" memory manager).
You have 2 choices:
1. re-configure with the configure switch --with-memory-manager=none
2. upgrade to the latest 7.1.2 beta (aside from a pending IMPI fix,
they're pretty stable): http://www.lam-mpi.org/beta/
On Nov 21, 2005, at 1:14 PM, Jeff Squyres wrote:
> Wait, I see the problem here... Based on the fact that you're not even
> getting in main, I went and looked closer at your config.log. It
> appears that somehow LAM is deciding that you should have the ptmalloc2
> memory manager instead of "none" (i.e., you should not be using a
> memory manager on any BSD).
>
> Can you run config.guess and send the output? We have a test in
> configure that should set anything with a config.guess output of *bsd*
> to use the "none" memory manager. Either our test is wrong, or somehow
> that default is being overridden.
>
>
> On Nov 21, 2005, at 12:31 PM, <v.kuznetsov_at_[hidden]> wrote:
>
>> Hi again,
>> 1. I compiled laminfo with debugging info:
>> g++ -g -O3 -o laminfo laminfo.o -pthread
>> ../../share/libmpi/.libs/libmpi.a
>> ../../share/liblam/.libs/liblam.a -lutil
>> 2. ran it with a breakpoint to call of first function:
>> less laminfo.cc:
>> // Initialize the argv parsing handle
>>
>> ad = ao_init();
>>
>>
>> ATB# gdb /usr/local/bin/laminfo
>> GNU gdb 4.18 (FreeBSD)
>> -------------
>>
>> (gdb) break ao_init
>> Breakpoint 1 at 0x8090bde
>> (gdb) run
>> Starting program: /usr/local/bin/laminfo
>> [hangs]...
>> ^C
>> Program received signal SIGINT, Interrupt.
>> 0x48197900 in __sys_poll () from /usr/lib/libc_r.so.4
>>
>> So it even don't comes to first function ao_init!
>> 3. I recompiled laminfo with one printf string before any action:
>> bool want_all = false;
>> printf("ffdfd\n");
>>
>> // Initialize the argv parsing handle
>>
>>
>> Ran it and it hangs too.
>> 4. simple "hello" C and c++ compiles and runs without problem.
>>
>> May be I simply gives you access to this freebsd box? If you send me
>> your
>> PGP key I'd be glad to make you an account to this box.
>> Thank you,
>> Vassili
>>
>>
>>> -----Original Message-----
>>> From: lam-bounces_at_[hidden] [mailto:lam-bounces_at_[hidden]] On
>>> Behalf
>>> Of Jeff Squyres
>>> Sent: Monday, November 21, 2005 7:04 PM
>>> To: General LAM/MPI mailing list
>>> Subject: Re: LAM: FW: FreeBSD MPI_INIT problem
>>>
>>> Based on your stack trace, it looks like something is getting
>>> corrupted. Unfortunately, I do not have access to any FreeBSD
>>> machines
>>> to test this myself. However, laminfo is clearly not getting very
>>> far
>>> -- can you step/next through from the beginning of the program to see
>>> where exactly it is failing?
>>>
>>
>>
>> _______________________________________________
>> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>>
>
> --
> {+} Jeff Squyres
> {+} The Open MPI Project
> {+} http://www.open-mpi.org/
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
--
{+} Jeff Squyres
{+} The Open MPI Project
{+} http://www.open-mpi.org/
|
