Stakeholder meeting notes – 2015/09/17

by Kyle Knoepfel in Events
Attendees:

Kurt Biery, Martin Frank, Patrick Gartung, Chris Jones, Tom Junk, Rob Kutschke, Adam Lyon, Gianluca Petrillo, Brian Rebel

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

Discussion of resolved issues:

Issue #7852: A module failing to put() a product it produces() should be an error

New FHiCL parameter that producer and filter configurations may use to disable the now-default behavior of throwing an exception whenever a product that has been declared via produces is not put onto the event:

moduleLabel: {
module_type: MyModule
errorOnFailureToPut: false
}

The FHiCL parameter is set to true by default. There is also a global flag which controls overall behavior: “services.scheduler.errorOnFailureToPut”.

If the global flag is true, the individual module instances can override the behavior by setting the local flag “moduleLabel.errorOnFailureToPut” to false, as shown above.
If the global flag is false, then any attempt to override the global flag at the per-module level will be ignored.

No change to user code/configuration is required to adopt the new default behavior.

Discussion of open issues:

Issue #10172: new product stack

There will be a new product stack for art 1.16 (see link above), including:

python v2_7_10
boost v1_59_0
mysql_client v5_6_26
root v5_34_32

Chris G. also recommended new versions of scons​​, and sqlite; Lynn recommended ifdhc. [Update: the stack has expanded a little more, and it may continue to do so until the release is cut.]

Issue #8653: ROOT6 migration

Paul reports that he has only 2 failing tests remaining (out of an original 250 or so). As part of the test suite, the artists will verify that ROOT5-generated art output files will be readable by ROOT6-compatible art. Chris J. reported that this was not a problem with CMS, except that the checksums are different.

Marc stated that we will need to build all the way through artdaq to make sure the migration is successful.

(Related discussions)

ROOT6 is not compatible with GCC 5 because of ABI naming incompatibilities. Unless there is a strong desire from the stakeholders, the artists are not pushing in the near future for a newer gcc.

Chris G. reports that ROOT5’s end of life is imminent.

Issue #7449: File-level data products (ResultsProducers)

Chris. G. mentioned that the Results-level product is nominally feature-complete, modulo testing and bug-fixing.

(Related discussions)

The module label associated with the Results-level product will have a pattern like OMlabel::RPModuleLabel. Brian had no concerns with that. [Update: since the double colon will cause problems with art::InputTag, a different delimiter will be used.]

The Results-level data product will be released in art 1.16, which will also include bug fixes. Marc asked if there were any objections to also including FHiCL configuration validation. None were voiced. An email will go out to the stakeholders asking for their opinions. By including FHiCL validation in this release, it:

– allows users to enable configuration validation/description for their own modules and services
– enables validation/description for art-provided services (not including messagefacility) and modules

For art-provided services and modules that are correctly configured, there will be no change in behavior. For configurations that provide extra/misspelled parameters, an exception will be thrown and art will shutdown gracefully.