[Mvapich-discuss] Segmentation fault if MPI_Finalize is called without freeing objects with attributes attached

[Subramoni, Hari]

Dear, Mose.

Glad to hear that the patch fixed the issues you were facing.

Thanks for the heads up. We will take a look at it and do the needful.

Best,
Hari.

-----Original Message-----
From: Giordano, Mose <m.giordano at ucl.ac.uk> 
Sent: Tuesday, January 17, 2023 11:19 AM
To: Subramoni, Hari <subramoni.1 at osu.edu>; Simon Byrne <simonbyrne at gmail.com>; mvapich-discuss at lists.osu.edu
Subject: Re: [Mvapich-discuss] Segmentation fault if MPI_Finalize is called without freeing objects with attributes attached

!-------------------------------------------------------------------|
  This Message Is From an External Sender
  This message came from outside your organization.
|-------------------------------------------------------------------!

Dear Hari,

I compiled MVAPICH 2.3.7 with your patch and I can confirm the test case provided by Simon doesn't result in a segmentation fault anymore.  The MPI.jl test suite doesn't result in any segmentation faults either.  Thanks a lot!

Tangentially related, I wanted to slightly speed up compilation of MVAPICH 2.3.7 and disabled Fortran and C++ bindings with `--disable-fortran --disable-cxx`, because for these tests I only need the C API, but this results in errors like

```
../../libtool: line 1734: no: command not found
../../libtool: line 1734: no: command not found
../../libtool: line 1734: no: command not found ```

during compilation.  The problem is the same as https://urldefense.com/v3/__https://github.com/pmodels/mpich/pull/4382__;!!KGKeukY!1go1_628ItRVkIYYHKlQN0hLYO1trw1ZDrhrI1Dj89uTy89kZsgz-D6ha8m3wfHKJ5DbIhtGVQUb3VypqZjZFTQ_$ : you shouldn't override CXX.  I see `configure.ac` has been already updated in MVAPICH 3 to match MPICH `configure.ac`, but if the 2.x series is still maintained you may want to apply the same patch there, otherwise `--disable-cxx` will be unusable.

Best,
Mosè