----------------------------------------------------------------------
                        KNOWN ISSUES
----------------------------------------------------------------------

### CH4

 * Build Issues
   - CH4 will not build with Solaris compilers.
 * Stability with large buffers
   - CH4 may suffer from stability issues when operating on large
     communication buffers (>= 2GB)

 * Dynamic process support
   - Requires fi_tagged capability in OFI netmod
   - Not supported in UCX netmod at this time

 * Test failures
   - CH4 will not currently pass 100% of the testsuite.
   - Test failures differ based on experimental setup and network
     support built in.

### OS/X

 * C++ bindings - Exception handling in C++ bindings is broken because
   of a bug in libc++ (see
   https://github.com/android-ndk/ndk/issues/289).  You can workaround
   this issue by explicitly linking your application to libstdc++
   instead (e.g., mpicxx -o foo foo.cxx -lstdc++).

 * GCC builds - When using Homebrew or Macports gcc (not the default
   gcc, which is just a symlink to clang), a bug in gcc
   (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=40838) causes stack
   alignment issues in some cases causing segmentation faults.  You
   can workaround this issue by building mpich with clang, or by
   passing a lower optimization level than the default (-O2).
   Currently, this bug only seems to show up when the
   --disable-error-checking flag is passed to configure.

### PGI compilers

 * ch3/nemesis will segfault during MPI_Init when built with recent
   versions of PGI compilers.

### PathScale compilers
 * Due to bugs in the PathScale compiler suite, some configurations of MPICH
   do not build correctly.
    - v5.0.1: When the --disable-shared configure option is passed to MPICH,
      applications will give a segfault.

    - v5.0.5: Unless you pass the --enable-fast=O0 configure flag to MPICH,
      applications will hang.

   See the following ticket for more information:

    https://trac.mpich.org/projects/mpich/ticket/2104

### Fine-grained thread safety

 * ch3:sock does not (and will not) support fine-grained threading.

 * MPI-IO APIs are not currently thread-safe when using fine-grained
   threading (--enable-thread-cs=per-object).

 * ch3:nemesis:tcp fine-grained threading is still experimental and may
   have correctness or performance issues.  Known correctness issues
   include dynamic process support and generalized request support.


### Lacking channel-specific features

 * ch3 does not presently support communication across heterogeneous
   platforms (e.g., a big-endian machine communicating with a
   little-endian machine).

 * Support for "external32" data representation is incomplete. This
   affects the MPI_Pack_external and MPI_Unpack_external routines, as
   well the external data representation capabilities of ROMIO.  In
   particular: noncontiguous user buffers could consume egregious
   amounts of memory in the MPI library and any types which vary in
   width between the native representation and the external32
   representation will likely cause corruption.  The following ticket
   contains some additional information:

     http://trac.mpich.org/projects/mpich/ticket/1754

 * ch3 has known problems in some cases when threading and dynamic
   processes are used together on communicators of size greater than
   one.


### Process Managers

 * Hydra has a bug related to stdin handling:

     https://trac.mpich.org/projects/mpich/ticket/1782


### Performance issues

 * SMP-aware collectives do not perform as well, in select cases, as
   non-SMP-aware collectives, e.g. MPI_Reduce with message sizes
   larger than 64KiB. These can be disabled by setting the environment
   variable MPIR_CVAR_ENABLE_SMP_COLLECTIVES to 0.

 * MPI_Irecv operations that are not explicitly completed before
   MPI_Finalize is called may fail to complete before MPI_Finalize
   returns, and thus never complete. Furthermore, any matching send
   operations may erroneously fail. By explicitly completed, we mean
   that the request associated with the operation is completed by one
   of the MPI_Test or MPI_Wait routines.