You should check that your path would also hit /usr/bin/mpiexec and not some other version of it
On Sep 17, 2014, at 4:01 PM, Nico Schlömer <nico.schloe...@gmail.com> wrote: > Hi all! > > Today, I observed a really funky behavior of my stock > ``` > $ mpiexec --version > mpiexec (OpenRTE) 1.6.5 > > Report bugs to http://www.open-mpi.org/community/help/ > ``` > on Ubuntu 14.04. When running one of my test codes with > ``` > $ mpiexec -n 2 ioTest > [...] > ``` > all is fine. If instead I use the full path of mpiexec, I get a > different behavior > ``` > $ /usr/bin/mpiexec -n 2 ioTest > [...] > (exception thrown) > ``` > I was puzzled, so skimmed the manpage and found that the `--prefix` > option might have something to do with it. I played around and got > ``` > $ /usr/bin/mpiexec --prefix . -n 2 ioTest > [fuji:21003] *** Process received signal *** > [fuji:21003] Signal: Segmentation fault (11) > [fuji:21003] Signal code: Address not mapped (1) > [fuji:21003] Failing at address: 0x1000000dd > [fuji:21003] [ 0] /lib/x86_64-linux-gnu/libpthread.so.0(+0x10340) > [0x7f12e4069340] > [fuji:21003] [ 1] /lib/x86_64-linux-gnu/libc.so.6(_IO_vfprintf+0x1d13) > [0x7f12e3cde8f3] > [fuji:21003] [ 2] > /lib/x86_64-linux-gnu/libc.so.6(__vasprintf_chk+0xb5) [0x7f12e3d9e035] > [fuji:21003] [ 3] > /usr/lib/libopen-rte.so.4(opal_show_help_vstring+0x343) > [0x7f12e43043e3] > [fuji:21003] [ 4] /usr/lib/libopen-rte.so.4(orte_show_help+0xaf) > [0x7f12e42a5faf] > [fuji:21003] [ 5] /usr/bin/mpiexec() [0x403ab3] > [fuji:21003] [ 6] /usr/bin/mpiexec() [0x40347d] > [fuji:21003] [ 7] > /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf5) > [0x7f12e3cb4ec5] > [fuji:21003] [ 8] /usr/bin/mpiexec() [0x403399] > [fuji:21003] *** End of error message *** > Segmentation fault (core dumped) > ``` > That's tough! > > Should I try to reproduce this with a more recent version? Any idea > what the reason for the different behavior of `mpiexec` and > `/usr/bin/mpiexec` might be? > > Cheers, > Nico > _______________________________________________ > users mailing list > us...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/users > Link to this post: > http://www.open-mpi.org/community/lists/users/2014/09/25338.php
