Issue 762

[vehre]

coverage on master

Summary of changes

Fix crash on certain platforms on finalize.

Rationale for changes

On finalize opencoarray was MPI_Win_detaching a management structure instead of previous Win_attached token. Detaching the token fixes the crash with openmpi. The testcase provided in the first commit does/can not really test the fix, because the tests do not check for crashing tests.

Additional info and certifications

This pull request (PR) is a:

• Bug fix
• Feature addition
• Other, Please describe:

I certify that

• I certify that:
  • I have reviewed and followed the contributing guidelines
  • I will wait at least 24 hours before self-approving the PR to give another
    OpenCoarrays developer a chance to review my proposed code
  • I have not introduced errant white space (no trailing white space or white space errors may
    be introduced)
  • I have added an explanation of what these changes do and why they should be included
  • I have checked to ensure there aren't other open Pull Requests for the same change
  • I have you written new tests for these changes
  • I have successfully tested these changes locally
  • I have commented any non-trivial, non-obvious code changes
  • The commits are logically atomic, self consistent and coherent
  • The commit messages follow best practices
  • Test coverage is maintained or increased after this is merged

Code coverage data

[everythingfunctional]

I've played around with this for a bit now, and can confirm it solves the crashing with openmpi. I have no idea why the change fixes it for openmpi without breaking mpich, or why it originally worked with mpich but not openmpi, but it works now. 🤷‍♂️

I also figured out why I was getting incorrect results with mpich. cafrun was still using mpiexec from openmpi, even though caf did link the executable to mpich. Perhaps something to look into 🤷‍♂️? I still think the new test is valuable, so no sense in getting rid of it.

Now the bad news. We still crash with Intel mpi. I'd say we should still merge this and call it a win, we just aren't done yet.

[vehre]

Intel crashing should be fixed by commit 9d4afcb. At least it does on my Fedora 35 Linux with Intel MPI 2021.6 .

[everythingfunctional]

I can confirm that this did solve the crashing with Intel MPI on Linux, and did improve the situation on Windows. Unfortunately, it does still crash on Windows, now with less severity it seems. The output from Windows is:

C:\Users\brad\Repositories\GitHub\sourceryinstitute>opencoarrays-install\bin\caf --show

C:\Users\brad\Repositories\GitHub\sourceryinstitute>"C:/Program Files/Git/usr/bin/bash.exe" "C:\Users\brad\Repositories\GitHub\sourceryinstitute\opencoarrays-install\bin\caf" --show
C:/Users/brad/gcc/bin/gfortran.exe -I/c/Users/brad/Repositories/GitHub/sourceryinstitute/opencoarrays-install/include/OpenCoarrays-2.10.0-14-g9d4afcb_GNU-12.1.0 -fcoarray=lib ${@} /c/Users/brad/Repositories/GitHub/sourceryinstitute/opencoarrays-install/lib/libcaf_mpi.a -pthread C:/Program Files (x86)/Intel/oneAPI/mpi/latest/lib/release/impi.lib

C:\Users\brad\Repositories\GitHub\sourceryinstitute>opencoarrays-install\bin\cafrun --show

C:\Users\brad\Repositories\GitHub\sourceryinstitute>"C:/Program Files/Git/usr/bin/bash.exe" "C:\Users\brad\Repositories\GitHub\sourceryinstitute\opencoarrays-install\bin\cafrun" --show
C:/Program Files (x86)/Intel/oneAPI/mpi/latest/bin/mpiexec.exe -n <number_of_images> /path/to/coarray_Fortran_program [arg4 [arg5 [...]]]

C:\Users\brad\Repositories\GitHub\sourceryinstitute>opencoarrays-install\bin\caf hello_coarrays.f90 -o hello_coarrays

C:\Users\brad\Repositories\GitHub\sourceryinstitute>"C:/Program Files/Git/usr/bin/bash.exe" "C:\Users\brad\Repositories\GitHub\sourceryinstitute\opencoarrays-install\bin\caf" hello_coarrays.f90 -o hello_coarrays

C:\Users\brad\Repositories\GitHub\sourceryinstitute>opencoarrays-install\bin\cafrun -n 4 .\hello_coarrays.exe

C:\Users\brad\Repositories\GitHub\sourceryinstitute>"C:/Program Files/Git/usr/bin/bash.exe" "C:\Users\brad\Repositories\GitHub\sourceryinstitute\opencoarrays-install\bin\cafrun" -n 4 .\hello_coarrays.exe
           1           1
           3           3
           4           4
           2           2

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   RANK 0 PID 6520 RUNNING AT BRADRICHARD5FC1
=   EXIT STATUS: -1073740940 (c0000374)
===================================================================================

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   RANK 1 PID 444 RUNNING AT BRADRICHARD5FC1
=   EXIT STATUS: -1 (ffffffff)
===================================================================================

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   RANK 2 PID 4796 RUNNING AT BRADRICHARD5FC1
=   EXIT STATUS: -1 (ffffffff)
===================================================================================

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   RANK 3 PID 8996 RUNNING AT BRADRICHARD5FC1
=   EXIT STATUS: -1 (ffffffff)
===================================================================================
Error: Command:
   `C:/Program Files (x86)/Intel/oneAPI/mpi/latest/bin/mpiexec.exe -n 4 .\hello_coarrays.exe`
failed to run.

I'm still of the opinion that if solving this last issue is much more effort, we can go ahead and merge this and solve the last thing as a separate PR. Up to you though.

[vehre]

I tried to debug this under Windows, but I see a different error. I see this error:

f:\dd\vctools\crt\crtw32\misc\dbgheap.c(1322) : Assertion failed: _CrtIsValidHeapPointer(pUserData)

This seems to be runtime related and I have no clue how to debug this on windows. So lets merge the existing fixes and if this is important do another round.