Stakeholder meeting notes – 2015/05/14

by Kyle Knoepfel

Kurt Biery, Tom Junk, Gianluca Petrillo, Chris Jones, Patrick Gartung, John Freeman, Andrei Gaponenko, Brian Rebel, Chris Backhouse, Rob Kutschke

artists: Chris Green, Jim Kowalkowski, Kyle Knoepfel, Marc Paterno


art 1.14.01

All versions have been distributed.

A known issue is that all modules now need to include the Boost_FILESYSTEM_LIBRARY and Boost_SYSTEM_LIBRARY libraries in the link list. In future releases of art, this will be automatically done for people who use the simple_plugin cetbuildtools macro.

Another known issue is that for users of the secondary file reading feature, a tree two generations deep yields a segmentation violation. This has been corrected and the fix will be included in the next art release.

Discussion of issues:

Issue #8602 – handling of signal values

Precedent in the past has been that issuing ‘Control-C’ yields a graceful shutdown and is a successful art process with return code 0. Consensus was reached among attendees that the default behavior will be to issue the corresponding return code to whenever (e.g.) Control-C is issued (not 0), although a command-line flag will allow users to retain current behavior. ‘Control-C’ still results in a graceful shutdown to the process.

Issue #8655 – Adjusting PROLOG variables in FHiCL document

Andrei discussed the issue: when modifying FHiCL, it’s easy to append. There is no sane way, however, to overwrite PROLOG variables. Mu2e wants the ability to change PROLOG variables from a non-interactive context. For PROLOG, last one wins. An adjustment to the behavior or syntax is requested to be able to override variables.

Significant discussion on this issue. In the end, the following options were introduced:

1) Reverse behavior of PROLOG parameters (first wins).
2) Append subsequent PROLOG statements to previous ones.
3) Introduce EPILOG statement which appears at end of file; the values inside the EPILOG override those in the PROLOG.
4) Provide art utility that would leave the current behavior and syntax as is, but append the provided PROLOG file to the other PROLOG files specified in the configuration file.

Decision to carry on discussion by email.

Discussion of resolved issues tabled until next week.


John Freeman reports erroneous user deprecation statement. This has been fixed for 1.14.

Marc mentioned that ROOT6 subtasks have been produced. Portability with Clang/Intel is not part of the same task. We now have ups product for ROOT6. Chris reports that we believe we can move from current forced loading of all found dictionaries to a regime where we can load one-by-one on demand. Users will need to produce rootmap files. For anyone using mrb or cetbuildtools, this producing rootmap files will be done automatically. Marc says that forwards compatibility (reading ROOT6 files with ROOT5 code) is not recommended and would be nontrivial to implement.