Issues with UCX

[dalcinl]

Thank you for taking the time to submit an issue!

Background information

What version of Open MPI are you using? (e.g., v3.0.5, v4.0.2, git branch name and hash, etc.)

master

Describe how Open MPI was installed (e.g., from a source/distribution tarball, from a git clone, from an operating system distribution package, etc.)

./configure --with-ofi --with-ucx ---with-pmix=internal -enable-debug --enable-mem-debug

If you are building/installing from a git clone, please copy-n-paste the output from git submodule status.

 9984334f91eadd17ebf3618b95d74fbca9c708fc 3rd-party/openpmix (v1.1.3-3208-g9984334f)
 b2d1226461f9be2689779a2f3b8503987559f69e 3rd-party/prrte (psrvr-v2.0.0rc1-4051-gb2d1226461)

Please describe the system on which you are running

• Operating system/version: Fedora 34
• Computer hardware: x86_64
• Network type: isolated

---

Details of the problem

I fresh run of mpi4py testsuite is printing the following malloc zero warnings and aborts after failed assertion:

...
malloc debug: Request for 0 bytes (osc_ucx_comm.c, 661)
malloc debug: Request for 0 bytes (osc_ucx_comm.c, 972)
malloc debug: Request for 0 bytes (osc_ucx_active_target.c, 228)
malloc debug: Request for 0 bytes (osc_ucx_active_target.c, 104)
...
osc_ucx_component.c:610: ompi_osc_ucx_win_attach: Assertion `insert_index >= 0 && (uint64_t)insert_index < module->state.dynamic_win_count' failed.

Runinng with OMPI_MCA_osc=sm, the assertion abort goes away, and my OSC tests complete successfully.

[devreal]

@janjust this seems to come from the dynamic windows fixes?

[janjust]

@devreal Thank you, I'll take a look. I actually discovered other issues in post/wait. I'll take a look at this asap.

[janjust]

Fixed with #10126

[gpaulsen]

Fixed in v5.0.x with #10138