Choose timezone
Your profile timezone:
ROOT Review planning
Present: the Root team, Andrea, Andreas, Pere, Marco, Peter, Bill (R), Paul (R)
Actions from last week:
ACTION: Valeri Check with Sebastien the status of the plugin manager and report
Actions:
ACTION Peter: Share the slides about root5/root6 on the weekly planning meeting
ACTION Fons: tag root6 for alice to prepare the grid distribution
ACTION Axel: help Valeri in laying down a strategy to get rid of the Reflex API in Atlas in ROOT5
ACTION Marco: (next or next2next week) report issues of LHCb tests
ACTION Lorenzo: Follow up on the failing histogram copy constructor and the missing TFitterMinuit.h header
Axel:
o Upgrade of the GetListOfXXX() method of TClass. New implementation is smarter, a more optimised lookup
o Vasil works on the unloading which is due for May since its expertise is needed there given the complexity of the issues and is not fundamental for the beta.
Peter:
o All tests work of root6 and root5
o Comparison of performances will be next week during the Alice week. Slides will be uploaded in the agenda of this meeting
o Grid large scale test will be started when a tag for root6 is available
o Small scale test done for analysis: successful
Valeri:
o Much time eaten by the S&C week
o Continue to use ROOT5 to eliminate the usage of Reflex
o The ideal procedure would be the LCHB's one: cook a stack which is Reflex free and use it with Root5 and Root6
- This also provides a reference for validation
Marco:
o Up to now the focus was more on Gaudi than the rest of the stack
o Some issues *maybe* linked to JIT-ting of templated functions
- This code is between Python and C++
- It's important but not a blocker for the beta (heavy interactivity in Python)
o Other error: histogram filled with very different number of entries in root5 and root6
o Started to look into the failing tests of LHCb:
- Maybe report next week or in 2 weeks
o Something changed with the policy of prinitng warnings in Python
- The warnings are due for example to the inclusion of Python.h after some other headers due to definition of constants
- They were not there yesterday and they are there today :(
o There is a general appreciation of the significant effort made by Marco in adapting LHCb software to root6
Bill:
o The CMS IB is up and running now
o All the dictionaries are generated and compile
o A list of items to be ironed out:
- C++ type is represented also with Reflex within CMSSW
- Properties of Reflex needs to be replaced by new TClass equivalent (these are already available in root)
- Conditions (Andreas is responsible for those)
- TFitterMinuit.h has disappeared in ROOT6
- A class was copy constructed correct in ROOT5 and ROOT6
Andrea:
o The TListOfXXX is important for PyCool: namespace issues are hitting
o C++ exceptions are not propagated to Python
- LHCb does not rely on this, but it's an important feature
- It is a fundamental issue in ROOT: this will be followed up closely as it is linked to the JIT engine
o Filed a bug about a different crash ROOT-5643
o JIT does not support inline asm: the issue does not seem to be high in the priority, it needs to be followed up anyway by Wim and Axel
Actions from last week:
ACTION: Valeri Check with Sebastien the status of the plugin manager and report
Actions:
ACTION Peter: Share the slides about root5/root6 on the weekly planning meeting
ACTION Fons: tag root6 for alice to prepare the grid distribution
ACTION Axel: help Valeri in laying down a strategy to get rid of the Reflex API in Atlas in ROOT5
ACTION Marco: (next or next2next week) report issues of LHCb tests
ACTION Lorenzo: Follow up on the failing histogram copy constructor and the missing TFitterMinuit.h header
Axel:
o Upgrade of the GetListOfXXX() method of TClass. New implementation is smarter, a more optimised lookup
o Vasil works on the unloading which is due for May since its expertise is needed there given the complexity of the issues and is not fundamental for the beta.
Peter:
o All tests work of root6 and root5
o Comparison of performances will be next week during the Alice week. Slides will be uploaded in the agenda of this meeting
o Grid large scale test will be started when a tag for root6 is available
o Small scale test done for analysis: successful
Valeri:
o Much time eaten by the S&C week
o Continue to use ROOT5 to eliminate the usage of Reflex
o The ideal procedure would be the LCHB's one: cook a stack which is Reflex free and use it with Root5 and Root6
- This also provides a reference for validation
Marco:
o Up to now the focus was more on Gaudi than the rest of the stack
o Some issues *maybe* linked to JIT-ting of templated functions
- This code is between Python and C++
- It's important but not a blocker for the beta (heavy interactivity in Python)
o Other error: histogram filled with very different number of entries in root5 and root6
o Started to look into the failing tests of LHCb:
- Maybe report next week or in 2 weeks
o Something changed with the policy of prinitng warnings in Python
- The warnings are due for example to the inclusion of Python.h after some other headers due to definition of constants
- They were not there yesterday and they are there today :(
o There is a general appreciation of the significant effort made by Marco in adapting LHCb software to root6
Bill:
o The CMS IB is up and running now
o All the dictionaries are generated and compile
o A list of items to be ironed out:
- C++ type is represented also with Reflex within CMSSW
- Properties of Reflex needs to be replaced by new TClass equivalent (these are already available in root)
- Conditions (Andreas is responsible for those)
- TFitterMinuit.h has disappeared in ROOT6
- A class was copy constructed correct in ROOT5 and ROOT6
Andrea:
o The TListOfXXX is important for PyCool: namespace issues are hitting
o C++ exceptions are not propagated to Python
- LHCb does not rely on this, but it's an important feature
- It is a fundamental issue in ROOT: this will be followed up closely as it is linked to the JIT engine
o Filed a bug about a different crash ROOT-5643
o JIT does not support inline asm: the issue does not seem to be high in the priority, it needs to be followed up anyway by Wim and Axel
There are minutes attached to this event.
Show them.