Lisandro Dalcin

Yes. `Request` is a bit special. For `Request` instances, `Wait()` or `Test()` methods implicitly work as a destructor. `Prequest` (subclass of `Request`) instances always require a `Free()` call. The case...

This is expected, as the test is creating sockets using `AF_INET` rather than `AF_INET6`. I wrote these tests long ago, circa 2014, just a couple years after IPV6 debut. As...

> then push a PR. Please make it against the `maint` branch. BTW, you may want to apply some other patches in the `maint` branch, particularly 9ccf55fb0a0218d3fcc5eb7292ed14117d781852.

@leofang and I pushed for Open MPI and MPICH to provide this API with uniform runtime semantics. @leofang has some preliminary code in his personal mpi4py fork, although it needs...

> It also appears to be supported by MPICH MPICH has more than one API, depending on support for CUDA/ZE/HIP. Not sure if Open MPI or MVPICH support anything else...

@jsquyres Looks like that's the case. The two build below show that the regression comes from 4896db17dda399c6c389408a7dad1395d3094521. * https://github.com/mpi4py/mpi4py-testing/runs/7706338665?check_suite_focus=true corresponds to 5c302accafaf03d00950d3444fc3d29314bb88b1 (good) * https://github.com/mpi4py/mpi4py-testing/runs/7706340330?check_suite_focus=true corresponds to 4896db17dda399c6c389408a7dad1395d3094521 (broken)

@awlauria Can you provide any ETA for you looking into this?

An additional pointer: using a local debug build, the issue seems to be happen only with `MPI_Comm_spawn_multiple()`. All of my tests involving `MPI_Comm_spawn()` are successful.

1. Your Python code is missing a `import numpy as np` statement. Of course, this is a trivial omission and not the reason of your issues. 2. You did not...

@jlmaccal Any update? Can we close this issue?