Stakeholder meeting notes – 2015/06/25

by Kyle Knoepfel
Attendees:

Martin Frank, Brian Rebel

artists: Lynn Garren, Chris Green, Kyle Knoepfel, Marc Paterno

art Version 1.14.03

Version 1.14.03 has been released, fixing bug #9193. The details of the bug are nontrivial. It was encountered whenever each of the following conditions was satisfied:
– any (sub)run products are set for delay reading
– the (sub)run products are not explicitly retrieved by the user during the job
– the (sub)run products are not dropped from input or output
– the (sub)run products, when requested for retrieval by art are not in the last input file

See 1.14.03 release notes.

Issues:

Bug #9283: Malformed SQLite database

Report from NOvA that a malformed SQLite database was being retrieved from a (supposedly) art-produced data-products file. The cause of the bug resulted from attempting to read an SQLite database from a ROOT input file that was produced by executing ‘hadd’ over several art output files. Since ‘hadd’ does not understand how to aggregate TKey objects that are SQLite databases, the resulting output file from ‘hadd’ was not readable as an art input file.

Brian requested putting a notice somewhere regarding this unintended consequence when running ‘hadd’ over art output files. Marc thought a reasonable way forward is to have art throw an exception if an SQLite database is not available in a RootInput file.

Feature #8655: Mu2e has requested that these FHiCL enhancements be released earlier than Alpha Centauri. Anticipate a 1.15.00 release sometime next week with these features. No behavior change for currently valid FHiCL files.

Feature #9255: Ordering of printed-out ParameterSets

The printed-out order of ParameterSets from config_dumper corresponds to the order of the ParameterSet IDs. It can be difficult to ‘diff’ configurations this way even if minor differences between ParameterSets exist. The proposal is to print out the configuration alphabetically according to key, rather than numerically according to ParameterSetID. The artists will need to determine for which release this feature will be slated.

Other:

Martin asked when the next qualifier bump will be. Up until now, there has been no expressed desire from any of the experiments to accommodate the gcc 5 series. Once we move to gcc 5.x, we will update the qualifiers. However, after the Alpha Centauri release, one of the next big issues is to support OSX native clang builds.