Welcomed Cristina

News:

Outlook issue leaving it overnight can't send emails.

Shift handover:

Olivier to Stefan
- Some issues with doc generation
- Crashing after 23 seconds now.
- Forum everything is addressed.

Meeting summary:

IO:
- Discusson about naming and especially a request from CMS. More or less reached an agreement but need input from Jacob at some point.
- FCC review last week. Close out, finalize a few open issues but everyone is aware and we have a path forward.

Statistical analysis meeting:
- No one connected.

PPP:
- We had an external report from mu2e analysis tools group. Developing tools in C++ and Python based on an old framework based on root. Shows how they use it (event loop version). They're happy to collaborate with us and eventually move to RDataFrame, there are some intermediate steps.

• PPP this week distributed Rdtaframe.

Planning/Stakeholder:
- Nothing to report

ML4EP:
- Overview of the current activity (NGT, Fast-simulation, machine learning inference) - last to last week.

LIM:
- There was nothing interesting.

Retreat:
- Cern ideasquare booked. Topics are very much free and up for discussion. There is no specific focus - it can be specific or general. There is a proposal for RNTuple documentation. It's one of the points that came out of the review.

• Technical discussions:

• About memory management: removing implicit memory management, there are some global collections of things in ROOT. Getting an overview of what should be managed should be discussed in the retreat.

• Proposal is to accumulate them and and a poll and we try to converge on lowly bounded number of items for this hackathon.

Round Table

Betrand:

Everything has to be in debug, if python needs to be in debug. 2000 tests in debug, debug mode is slow, huge and slow.

Lorenzo:
Some test failing (with CLAD): Might be minuit/CLAD - needs to be further investigated. Close this issues on minuit2.

Marta:
Problem: Things work in multi-threaded but not single threaded

Vincenzo:
The difference between single and multithreaded might just be an ordering problem. Need to dig a bit deeper into this.

Philippe:

• Got frustrating distracted by build failures - not propagating libx and stdlib to LLVM; inconsistent and broken.
• Work with Vincenzo on load entry - issues with changing cursor - fixed that.
• CMS is playing the as-needed flag for their build. Error: consequence on their builds because they also have shared libraries containing ROOT dictionaries. Changed ordering of things and led to discovering some awkwardness in the code. Disable the as-needed for the shared library with some special treatment in CMake.