WeeklyTelcon_20170516

Open MPI Weekly Telcon

We still have some ARM issue
- What does our distro-repackagers for example Fedora / Debian do?
- If distro repackagers see that 32bit isn't working, do they still release 64bit?
Still planning to ship v2.0.3 despite Pasha and Oreon's Issue 2643
Open HPC is on Open MPI 1.10.6.
- Ralph will ask if they can upgrade.
Current schedule is start RCs May 23rd (or so).

v2.1.1 Shipped last week!!!
Howard has begun merging stuff into v2.x
Question on PR3427 - remove ugenie ___ call. Work in Progress - Nathan will get to next week. Not a current bug.
Issue 3442 - 32bit builds are busted, probably affects v2.1.x also.

PMIx update:
- Cross version update is out there, but not working correctly. No shmem either.
- what's IN, seems to be working well other than dynamics (due to back end mapping system).
- one thing that needs to be done (internal stuff is looking okay)
- Need someone to look at external PMIx v1.2.2 and run with SLURM.
  - Artem reported orte integration in external PMIx is broken. Issue PMIx 3200 -not resolved.
  - Nathan could look at.
- Could we NAME PMIx something? Yes, quite possibly, Ralph will talk with PMIx community so that Open MPI v3.0 can go out with SOME PMIx versioned quantitity... possibly PMIx v2.0.0 (without compatibility).
Brian would like to do a Release Canidate next week.
- Howard would like to push the release canidate back a bit more.
- Want to do due-dilligence to ensure that all bugfixes in v2.x is in v3.0.
PR on master 3441 - slightly disturbing, and had to do some other changes to get it to merge into v3.0.
Brian has NOT yet done the branch renaming thing. We still see v3.x not v3.0.x
- Mohan will mention this to Brian, but he's out this week.

Absoft - still seeing Make Check fail - Issue 3522 (George)
- Giles is saying we should not make something const. George will look at it.
Putting asside OSHMEM errors.
- All failures seem to fall into dynamic and one-sided.
Geoffroy Valle - found a bug in PGI. He's starting to push master more. Running tests and successfully pushing back.

A bunch of PRs put in last week by Intel. Things are running well enough to begin pulling in Giles ini files.
Intel has an intern who's helping.
Ralph wanted to write a python script to read the perl ini files.
Made a change to ONLY allow Open MPI community members trigger build / test of PRs to prevent possible malicious commit PRs.

This week re-reiterated Security FYI.
- When you create a PR from a fork to a repo, GitHub opens up the branch permissions for all folks with write access to the REPO.
  - Anyone can even rebase your private fork's branch.
- GitHub assumes that projects have code-czars "maintainers" who may want to modify your private forks.
- When you create a PR, there is a check-box off to the side "allow edits by maintainers to modify PR".
  - Most users assume that this check-box only refer to PR title, and comments, not actual Content.
- As a Policy we agree that good community behavior would be to NOT do this, and only talk to the person before changing anything in their fork.
- Before this announcement, we thought that forks were always private... but not once you create a PR.
Request for folks to conclude discussion about PR2941
- Jeff will refresh and comment on.
Face2Face Meeting-2017-07
- Date: July 11-13 (9am Tuesday - noon on Thursday.
- Cisco has booked space in Chicago.
  - Cisco has reserved some space right next to O Hare (can get shuttle to hotel).
    - we have met there before.
  - Jeff will come in Monday evening.

Geoffroy Vallee - Starting to run MTT Open MPI v3.x and Master on Oakridge systems.
- Running into a few issues with PGI and LSF.
- Josh Hursey (IBM) is helping.