exported symbol name pollution test

[markalle]

This runs from inside a rank and uses ldd on itself to figure out what
libraries to examine, then nm on those MPI libraries to look for symbols
without some accepted OMPI prefix.

Signed-off-by: Mark Allen [email protected]

[gpaulsen]

@jjhursey Mark was asking if this will become part of Open-MPI mtt tests (or CI tests) once this gets merged. I'm not sure how to determine the answer to that, do you know?

[jjhursey]

I don't think this will cause MTT or CI to automatically pick it up.

For MTT we would need to add a new test suite for it since I don't think we have one setup for the ompi-tests-public yet. That's pretty easy to do though. Once this is merged in then I can show you how to do that.

If you want it in CI then we would need to talk about that a bit. We can probably add it to the OMPI AWS CI setup. It would be fairly easy to run in a container as a separate CI task, or just tack it onto the end of one of the existing builds.

[jjhursey]

Can you add a README.md at the package/ level that shows how to run this test? With some notes on expected output for both a successful run and a run where it fails?

That'll help when someone picks it up, and when we integrate it into MTT and/or CI.

[markalle]

Added a README.md

[markalle]

FWIW Geoffrey did point me at
https://www.mail-archive.com/[email protected]/msg21174.html
open-mpi/ompi#8132
and I suspect pulling the MCAs into the main libs will expose another pile of exported symbols without any OMPI-related prefixes in front of them.

[jjhursey]

There is still discussion happening on open-mpi/ompi#8132 before it is ready to merge, but (last I heard) the general feeling is that it will make it into the v5 stream. I think a tool like this will become more important in that case since we want to keep the component's exported symbols in the proper namespace. So this test will at least be a warning sign for us.

[markalle]

Minore repush: added "libnbc_" and "tm_" as accepted prefixes. I don't necessarily like "tm_" as a prefix, but since its usage appears to spread over into the prrte repo I wasn't as inclined to add an "ompi_" in front of it to fix it, but rather at least for now accept "tm_" as one of our recognized prefixes.

[markalle]

Back when I first made this there was some discussion about how best to fix symbol pollution, but I don't think that affects the testcase here. Regardless of how we fix it, I believe this is a correct testcase to identify when we've missed something. The most likely changes I could see needed is if someone objects to the "legal prefixes" list and wants it to be smaller.