The Open MPI Team, representing a consortium of research, academic, and industry partners, is pleased to announce the release of Open MPI version 1.4.2. This release is mainly a bug fix release over the v1.4.1 release. We strongly recommend that all users upgrade to version 1.4.2 if possible.
Version 1.4.2 can be downloaded from the main Open MPI web site or any of its mirrors (mirrors will be updating shortly). Here is a list of changes in v1.4.2 as compared to v1.4.1 - Fixed problem when running in heterogeneous environments. Thanks to Timur Magomedov for helping to track down this issue. - Update LSF support to ensure that the path is passed correctly. Thanks to Teng Lin for submitting a patch. - Fixed some miscellaneous oversubscription detection bugs. - IBM re-licensed its LoadLeveler code to be BSD-compliant. - Various OpenBSD and NetBSD build and run-time fixes. Many thanks to the OpenBSD community for their time, expertise, and patience getting these fixes incorporated into Open MPI's main line. - Various fixes for multithreading deadlocks, race conditions, and other nefarious things. - Fixed ROMIO's handling of "nearly" contiguous issues (e.g., with non-zero true_lb). Thanks for Pascal Deveze for the patch. - Bunches of Windows build fixes. Many thanks to several Windows users for their help in improving our support on Windows. - Now allow the graceful failover from MTLs to BTLs if no MTLs can initialize successfully. - Added "clobber" information to various atomic operations, fixing erroneous behavior in some newer versions of the GNU compiler suite. - Update various iWARP and InfiniBand device specifications in the OpenFabrics .ini support file. - Fix the use of hostfiles when a username is supplied. - Various fixes for rankfile support. - Updated the internal version of VampirTrace to 5.4.12. - Fixed OS X TCP wireup issues having to do with IPv4/IPv6 confusion (see https://svn.open-mpi.org/trac/ompi/changeset/22788 for more details). - Fixed some problems in processor affinity support, including when there are "holes" in the processor namespace (e.g., offline processors). - Ensure that Open MPI's "session directory" (usually located in /tmp) is cleaned up after process termination. - Fixed some problems with the collective "hierarch" implementation that could occur in some obscure conditions. - Various MPI_REQUEST_NULL, API parameter checking, and attribute error handling fixes. Thanks to Lisandro Dalc√≠n for reporting the issues. - Fix case where MPI_GATHER erroneously used datatypes on non-root nodes. Thanks to Michael Hofmann for investigating the issue. - Patched ROMIO support for PVFS2 > v2.7 (patch taken from MPICH2 version of ROMIO). - Fixed "mpirun --report-bindings" behavior when used with mpi_paffinity_alone=1. Also fixed mpi_paffinity_alone=1 behavior with non-MPI applications. Thanks to Brice Goglin for noticing the problem. - Ensure that all OpenFabrics devices have compatible receive_queues specifications before allowing them to communicate. See the lengthy comment in https://svn.open-mpi.org/trac/ompi/changeset/22592 for more details. - Fix some issues with checkpoint/restart. - Improve the pre-MPI_INIT/post-MPI_FINALIZE error messages. - Ensure that loopback addresses are never advertised to peer processes for RDMA/OpenFabrics support. - Fixed a CSUM PML false positive. - Various fixes for Catamount support. - Minor update to wrapper compilers in how user-specific argv is ordered on the final command line. Thanks to Jed Brown for the suggestions. - Removed flex.exe binary from Open MPI tarballs; now generate flex code from a newer (Windows-friendly) flex when we make official tarballs.