Migration to SL5 James Bellinger University of Wisconsin at Madison 21-Apr
Milestone Announced releases m, 6.1.4mc.m, and m last week Minor glitches but everything seems to be there Some packages (JointPhysics) still needed fiddling Well done 21-Apr-2010 Migration to SL5
development – Not an ordinary release – No validation: packaging and compilation only – Which external packages should be default? – Validation required Uses root v5_14_00g –q GCC_3_4_3 – Event display gripes but works – Are root files interoperable with old version? – Something is screwed up with the distribution New release 21-Apr-2010 What’s Next?
Need to be able to compile/link everything Not sure we need to have binaries distributed – Not going to be many customers for pulling a development distribution in the first place Leave development-lite structure alone? 21-Apr-2010 development
10K test OK for verifying code – We already verified the system libraries Verifying new root may need more effort: not a simple matrix – Write 616/SL3 read 701/SL5 and rewrite and compare – Re-read 616/SL3 and rewrite and compare » Re-read 701/SL4 and compare – Trying to find out what was already done Murat says they’ve been running successfully, doesn’t remember validation procedures 21-Apr m Validation
Reprocessing may be in the cards – Claim of 60% more top with modified b-tagging! Safe to bet that new compilers will call errors what are now just warnings – Extensive minor changes needed, probably not automatable. Unified Production/MC/ntuple release makes maintenance easier Probably follow-on is better than Apr-2010 New Release Desirable
What next? The less we support, the better? Can we use native – python? – perl? need to maintain oracle interface, so probably not – dddebugger? – gtools? bison and f2c don’t seem to bundle with SL5, rest do – gsl? – 21-Apr-2010 Offloading products?