ROOT Team Meeting
ROOT Meeting 6/5
News
DP: Last week we released RC 6.32, it went relatively smooth. We did not receive feedback but all relevant people got notified. More on this during the status update.
DP: We have a new team member, Giacomo, working on RNTuple and I/O supervised by Jakob.
Shift handover
DP: Had a few days off, VEP took over for those days. Relatively quiet with a few failures. Badly designed TMVA test, fixed by LM. We have instabilities in Windows. Everyone relevant has been notified.
DP: For the forum, quick response this week. Congrats and keep it up.
DP: For the PoW, we are still at 20%, not much movement compared to last week.
DP: RE issues, we are closing Jira issue
Meeting summaries
RooFit
JR: Shorter meeting (many people on vacation), introduced to some students working on Clad
I/O
VEP: Quick roundtable, no main items on the agenda. Major item was introducing Jack.
DP: Did you discuss the 1GB limit?
VEP: Jack and I discussed the RNTuple side, but it didn't come up in the meeting.
LIM
DP: Anything important for 6.32?
VEP: Trying to resurrect the Conda builds, getting close. I want to address the failures on Ubuntu 22, related to Python GC.
DP: I have some changes related to Fedora and Ubuntu, let's align them.
Topics
Release binaries builtins
DP: ROOT depends on several externals. The philosophy lately is to depend on the OS/package distribution system as much as possible instead of using builtins. Is this the right way forward? Right now we are moving to a situation where the binaries depend a lot on the external environment. This might become an issue for users of clusters. Maybe we can find a compromise so we have standalone binaries ready to distribute. E.g. last release we had a very vocal complaint on the forum regarding missing packages.
JR: I think if the communication in your institute is reasonable you should be able to request the packages from the sysadmin.
JB: Apparently these in these situations CVMFS is also not available?
DP: Sometimes emotional/political arguments weigh more than technical ones.
PC: Users don't always have permissions to install ROOT/dependencies to ROOT.
DP: We can be more explicit in the install instructions for use with CVMFS.
JB: This is there already.
JH: If a group uses a centrally managed cluster it's not as easy to go to the admin, it requires going to the official support channels. The question is do we expect users to link prebuild binaries to distribution packages?
PC: We have to not lose track of the fact that we choose a minimal build to improve security (i.e. using the latest versions of dependencies, automatic updates). More builtins means we have to be way more proactive.
DP: That's a very good argument.
JR: To illustrate, we have the OpenSSL EOL issue right now. On another note, we would also have to test builtins in the CI, but currently we disable them as much as possible. More builtins means more CI configurations.
CHEP
DP: Everything is in the hands of Gerri and Predrag. They will come back to us and tell us who can submit. I will check with Gerri who should actually submit.
Status updates
VEP: See status update doc
JH: RE the conda builds: /dev/urandom is not cryptographically secure IIRC.
JB: It's good enough.
VEP: For context, this is pertaining to the web graphics and only used for extremely old systems. Unfortunately this includes CC7, which here still has a big user base. There is an alternative, but will take a big effort and for what concerns 6.32.00 this is good enough.
JB: It's true that /dev/random is strictly better, but only for very esoteric use cases.
PC: In the upgrade to LLVM16 we fixed a bug that affected CMS in a surprising way. In a use case they had an expression parser and they got the type of an argument that was an enum tied to a using statement. They got the desugared name, instead of the spelled name with the upgrade, which broke their system. The issue that's still left is that you cannot get a class/enum that has been defined with the using directive. There is an issue open on GH.
SL: Some WebGUI/JSROOT work, found and fixed some issues.
DP: Can we add these as a test?
SL: I got conformation that I can go to CHEP, will probably submit two abstracts (WebGUI + JSROOT).
MD: See status update doc
LM: See status update doc
DP: What will the GSOC student do?
LM: They will add several missing operators
JR: See status update doc
JR: For the nix package, interesting to mention is they don't use the builtin LLVM, similar to Conda. They do use builtin Clang.
DP: This is an interesting use case we can learn and benefit from.
JH: See status update doc
VEP: For the artifact, what do they require?
JH: An archive of the benchmark sources. I asked if I need to include ROOT as well since it's already OS. Also need to provide a document with run instructions. If the committee is able to reproduce qualitatively, we get a badge on the first page of the paper.
JH: See status update doc
JB: See status update doc
GP: See status update doc
FdG: See status update doc
DVS: See status update doc
DVS: Worked with PC on the using issue
DP: I see you have a few PRs fully green and ready to be merged, if there's anything blocking their merger let us know.
DP: See status update doc
DP: Won't be here next Monday, I will handover the chair of the meeting to someone else.
AD: Working on the JSROOT testing suite. Made it possible to compare JSON files. Next up SVG as well. Also working on the automation of these tests.
KP: Discussion with VEP for the RBatchGenerator. The first option didn't work, so started working on the alternative.
DP: Let's fix a PPP presentation date.
JR: Related to the point of DP regarding enabling Fortran. What else should we enable in the CI?
DP: Let's not discuss this now, but we can set up a one-on-one to converge and then report during a future meeting.