Stakeholder meeting notes – 2015/10/01

by Kyle Knoepfel in Events

Chris Backhouse, Tom Junk, Rob Kutschke, Brian Rebel

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

art 1.16.00 – was released for SLF6, but there will not be a release for OSX due to a change going from XCode 6 to XCode 7, which now uses the clang assembler instead of the GNU assembler. This created problems for gcc 4.9.2 that were resolved in gcc 4.9.3. There will thus be a new compiler version used in building the OSX-supported art releases (e9 = gcc 4.9.3). The 1.16.00 release will be built only with e7.

art 1.16.01 – included fix due to failing test on OSX, and is built with e9. Lynn would prefer to build e7 only on request.

Discussion about release features

Only functional difference between 1.16.00/1.16.01 is the compiler version.

There was a discussion of the product stack – scons 2.4.0 was omitted from the list, but included in the bundles.

Chris G. also went through the list of new features, as discussed in the release notes here (1.16.00) and here (1.16.01). The primary feature is the NOvA-requested Results-level product. Also included is the FHiCL configuration validation and description, and sundry other features.

There are a few breaking changes related to deprecated functions—either having changed names or signatures. In addition to the breaking changes discussed in the release notes, the write* virtual functions of the OutputModule base class now have non-const reference arguments, which is unlikely to affect anyone.

Compiler support

Discussion about whether to continue with e7 series on Linux. There seemed to be general consensus that moving solely to e9 builds for 1.17 and beyond is fine. Brian would ask Martin Frank if switching to e9 only would cause issues with NOvA DDT.

Discussion about dropping products on input

Chris B. asked about whether it’s possible to drop products on input. It is if one specifies ’inputCommands: […]’ in the source block of the user’s FHiCL file, analogous to the outputCommands invocation that can be used for output streams. It seems, however, that maybe an event.getBySelector call could be more appropriate, where a customized selector can be provided.


Chris G. mentioned that we might be able to release 1.17 as early as next week.

Marc brought up that although the artists have not been pushing for GCC 5, the newly released Guideline Support Library for C++, which we would like to use, is not supported using GCC 4.9. There are difficulties with ROOT accommodating GCC 5 due to ABI changes. Marc is prodding ROOT, however, to fix them.