Choose timezone
Your profile timezone:
ROOT Project Meeting
ROOT Team Meeting
Zoom Meeting ID
63815034196
Host
Danilo Piparo
Alternative host
Philippe Canal
Useful links
Join via phone
Zoom URL
Shift handover
Jack -> Danilo
Danilo: shift went fairly well. PHP issues meant we had to close the forum for roughly one day. CI looks overall fairly green.
Issues & PRs
Danilo: "I closed" 6 issues with one PR showing reproducers for each of the respective issues. All of them were actually fixed by the migration to cppjit, so actually thanks to Aaron for all the good work!
Danilo: if there are any PRs you want to get merged before 6.42 add it to the github milestone so we can keep track of them together.
Important dates
Summaries
I/O
- Vincenzo: during the I/O meeting we thought it would be good to have a pre-discussion among I/O devs only.
PPP
No PPP last week, next one on November 5th
Python/C++ bindings
Aaron: we've seen a presentation by external contributors to cppjit with three different feature proposals
Vincenzo: I was curious about the one related to debugging, when debugging RDF Python applications I use CLING_DEBUG=1 with gdb, if there would be an even nicer way that integrates Python and C++ better I would definitely be willing to try it.
Aaron: this feature is slightly different than the one enabled by CLING_DEBUG.
Liaisons
ATLAS
- Florine: This Thursday Sanjiban will give a talk at the ATLAS S&C about SOFIE
- Florine: There was a bug report by ATLAS related to hadd. PR from external contributor with a proposed fix. Vincenzo and I took a look and validated with more tests. We would appreciate Philippe's review
LHCb
- Silia: will give a talk to LHCb on November 11th. Followup to the talk at the workshop in the beginning of the year. They shared a report that highlighted concretely actions for LHCb to collaborate with ROOT.
- Danilo: that's nice, we should communicate even more clearly now with experiments about ROOT 7 given that it's approaching.
FCC
- Juan is organising an in-person key4hep workshop at CERN in December. If there's something we want to do with FCC we can submit as discussion topics.
CMS
- Vincenzo: this wednesday CAT meeting with presentation of an analysis framework by the CERN CMS group using RDF. This is known, but now they want to have it as an official CAT framework. I will see if there are any major missing RDF features for them to work smoothly.
SOFIE round table
Round table
Vincenzo
-
Fixed an RDF issue, user wants backport to 6.32 without saying why. What should we do?
- Danilo: I can send an email to this person to show goodwill
-
Rate of PRs unsustainable. Am I the only one? I have a proposal to limit number of PRs.
- Aaron: I'm kind of co-authoring this proposal given we were discussing with Vincenzo on Friday.
- Danilo: I think no one will complain if you leave some PRs open, given we are all very loaded.
- Aaron: I think it's a strong case for AGENTS.md. The agents will read this document.
- Florine: it feels like all these PRs are from people who want to see their github PRs grow. What I think will stop is we write this in CONTRIBUTING.md, revisit PRs and close whenever we need to
- Aaron: But are you assuming people will read CONTRIBUTING.md?
- Florine: actually no, I think we should just punish people who obviously do not respect the rules.
- Aaron: to go against AGENTS.md, the user would need to make this explicit to their agent.
- Florine: But that's just validating the workflow of user and agent, they still won't read our explicit rules.
- Jakob: I find we can't completely ignore these contributions. Sometimes, they may be actually useful.
- Vincenzo: Yes, that's why I think throttling the number of open PRs is a viable mitigation.
- Florine: I think this is why we should close malicious PRs, take away satisfaction from these users.
ACTION (for Vincenzo): I will drive the proposal in view of a final decision to be taken next ROOT team meeting.
- About this PR https://github.com/root-project/root/pull/23791 that Florine and I worked on to validate. A word of caution for reviewers: the PR description is hallucinated
- Philippe: why mention it now rather than in the PR?
- Vincenzo: I wanted to bring up the point for discussion in the team. This PR is about fixing a critical I/O bug for an experiment, and it stems from a pure stroke of luck. We hope that with enough validation and coverage from our side we'll be sure this is the right fix, but it could have gone the opposite way. Actually, also thinking about Florine's point, I've been pondering whether we should just have closed the PR and reopened it ourselves, removing credit to the current author given that it's clear they didn't even know what they were doing.
Stephan
-
Stephan: Back from the school in Mexico. Varied audience, some people have never done HEP analysis.
-
Stephan: I was able to do 2 advanced sessions, I gave them options. What they chose was advanced RDF (I think particularly because of the ML data loader) and ROOT from C++.
-
Stephan: Anecdote about ML data loader: some postdocs in the audience saw the data loader and they saw immediate value with this. It seems like people see see value in this.
Silia
Philippe
-
Philippe: Attended HEP-CCE meeting. Major theme was how to address and harness AI and LLMs. It was clear that RNTuple is a corner stone for other tools/processes being built on top of it. Lot of interest for the RDF data loader.
-
Vincenzo: scope of lossy compression discussion?
- Philippe: no concrete plan, except for "planning to plan".
-
Philippe: SLAC joined HEP-CCE. In particular to involve XRootD.
-
Philippe: there's a rumor. Nick Smith and friends are investigating how to replace XRootD just with https protocol.
-
Philippe: there was also lot of interest on Samantha Abbot's presentation about the work (carried with Nick Smith) on RNTuple reading from object store.
Martin
- I'm using pip install ROOT to run my analysis on a cluster, it works!
Koen
Florine
- Florine: RNTupleProcessor out of experimental now!
Danilo
Bertrand
- I wanted to create a new Windows VM but noticed we are out of memory quota on openstack
- Danilo: I will open a ticket and request more quota
Aaron
- Vincenzo: just wanted to mention that tests that check that memory does not increase are inherently quite unreliable because of the vast amount of platforms and machines we test on.
*
There are minutes attached to this event.
Show them.