search

codes-org / codes

The Co-Design of Exascale Storage Architectures (CODES) simulation framework builds upon the ROSS parallel discrete event simulation engine to provide high-performance simulation utilities and models for building scalable distributed systems simulations
Other
40 stars 16 forks source link

Failed checks running 'make check' on Mac (Imported #166) #166

Closed nmcglo closed 5 years ago

nmcglo commented 5 years ago

Original Issue Author: Giorgis Georgakoudis Original Issue ID: 166 Original Issue URL: https://xgitlab.cels.anl.gov/codes/codes/issues/166

Running 'make check' for CODES latest version (f28191b7c11bb0a64b6e803587a00de17fee4784), using ROSS latest version (17e9d7b455be32fc2b085f38d9a5b3cc30c37101), results in 2 failed tests:

============================================================================
Testsuite summary for codes 1.0.0
============================================================================
# TOTAL: 22
# PASS:  20
# SKIP:  0
# XFAIL: 0
# FAIL:  2
# XPASS: 0
# ERROR: 0
============================================================================
See ./test-suite.log
Please report to http://trac.mcs.anl.gov/projects/codes/newticket
============================================================================
make[3]: *** [test-suite.log] Error 1
make[2]: *** [check-TESTS] Error 2
make[1]: *** [check-am] Error 2
make: *** [check] Error 2

I am attaching the test-suite.log

Thanks, Giorgis

test-suite.log

nmcglo commented 5 years ago

Caitlin Ross:

I can't replicate this issue, but it looks like it's specifically a Mac issue, which I don't use for development. Though I believe @neil says he hasn't had issues on a mac. Also, I think @mubarak develops on a Mac and I'm assuming she hasn't had issues either.

It doesn't really make sense why it can't find MPI_COMM_ROSS for these two test programs (which don't even use MPI_COMM_ROSS in the first place), but has no problem finding it for any of the others. I suggest deleting your build directories for both ROSS and CODES and rebuilding both. Wouldn't be the first time doing that has fixed a weird issue.

nmcglo commented 5 years ago

Caitlin Ross:

Hmm okay. We typically recommend to use MPICH, though I've used OpenMPI without issues too. Actually I just tested with OpenMPI 3.1.3 and GCC 7.3.0 and all tests pass, though this was on a Linux system.

nmcglo commented 5 years ago

Misbah Mubarak:

Did we have a closure on this one?

nmcglo commented 5 years ago

Caitlin Ross:

Not sure. It appears to be only a Mac issue (as all tests pass when using OpenMPI on a Linux system), so I think unless there's more information, we can probably close the issue.

nmcglo commented 5 years ago

Giorgis Georgakoudis:

Thanks, @caitlinross. I've investigated more. This issues exists when building ROSS/CODES with OpenMPI. For some reason the symbol MPI_COMM_ROSS is not exported and the linker fails to find it for those tests. I've tested with GCC and Apple Clang builds and it appears in both. If other mac users can replicate the issue that'd verify the problem.

nmcglo commented 5 years ago

Misbah Mubarak:

closed

nmcglo commented 5 years ago

Giorgis Georgakoudis:

changed the description

nmcglo commented 5 years ago

Giorgis Georgakoudis:

changed the description