Present: Laura Fields, Lynn Garren, Herbert Greenlee, Alex Himmel, Thomas Junk, Jim Kowalkowski, Jonathan Paley, Marc Paterno, Gianluca Petrillo, Ruth Pordes, Brian Rebel, Saba Sehrish, Erica Snider, Tingjun Yang, Joseph Zennamo
Remote: Vito Di Benedetto, Robert Sulej
Deployment plan for new CI test suite [Vito Di Benedetto]
-
code in branch feature/vdb_ci_regression_test_suite
-
[Marc Paterno] is there a back-compatibility check for schema evolution? A: not yet
-
[Erica Snider] it is in the requirements to have such a feature
-
needs to be tested in Jenkins
-
test "failures" are observed caused by the appearance of new data products respect to the reference
-
developing tests to use alternative art version
-
should the scripts (on slide 8) live in lar_ci? will discuss; [Lynn Garren] it depends on the scripts themselves
-
some "unit tests" are very long; should not they be C.I. tests? [Lynn Garren] they catch a lot of things that should be covered by unit tests, but haven't been so far
-
[Erica Snider] asked for requests or comments on the system before it is deployed: no comment received
Conclusions:
-
the location of the scripts (slide 8) has to be decided, according to their content
-
the long unit tests need to be reorganized
-
last call for comments before deployment happens, by e-mail
LArSoft status report [Erica Snider]
-
[Thomas Junk] can larsoft_data be recovered if removed from /grid/fermiapp? A: they are in scisoft
-
documentation is going to be stored in INSPIRE
-
[Alex Himmel] flow of the work to document a module: first go on with the normal experiment procedures (including publication in experiment DocDB), then there will be a script to publish it in INSPIRE; who runs this not-yet-existing scripts
-
[Thomas Junk] we have different formats than PDF. Can we upload those too? [Erica Snider] likely; we should make sure that our example suite samples them
-
[Gianluca Petrillo] new version of art introduces a potentially breaking change; details are in the LArSoft breaking changes web page
Architecture Review Status Report [Gianluca Petrillo]
-
[Jonathan Paley] what if the delivery is missed? what is driving this schedule?
-
[Gianluca Petrillo] at least good part of this work will facilitate the next phase, and it will need to be somehow performed any way
-
[Erica Snider] this will set an example for new code
Full reconstruction chain for DUNE FD [Robert Sulej]
-
[Erica Snider] any chance of studying no wrap? A: [Thomas Junk] DUNE is more worried about knowing precisely the efficiency than to maximise it; no-wrapping configurations can be simulated by cheating on the disambiguation to have it at 100% efficiency
-
[Tingjun Yang] TPC size is not going to be changed (due to physical constraints of experiment site)
-
the smallest wire angle under study still has all U/V wires with only one crossing per collection wire per TPC
LArSoft services factorization [Jonathan Paley]
-
[Brian Rebel] what is the trailing "
I" in the service name?
-
Erica Snider wants to kill it
-
Jonathan Paley does not care
-
Brian Rebel would rather see it away, but it's fine to discuss it off line
-
in general, survey proves most people to be softly opinionated, but preference is for removal
-
name convention (non-interface based services use the "interface" names):
|
Service type |
Name example (on DetectorProperties) |
|
provider interface |
DetectorProperties |
|
service interface |
DetectorPropertiesService |
|
provider implementations |
e.g. DetectorPropertiesSimple |
|
service implementation |
e.g. DetectorPropertiesSimpleService |
-
everything that depends on electric field has been moved to
DetectorProperties
-
[Herbert Greenlee] people do not like interface changes, especially for FHiCL file changes
-
can we provide tools to make it easier to update?
-
[Herbert Greenlee] scintillation model will depend on electric field; does his set of changes make it impossible to have it?
-
[Brian Rebel] do we really need a LArProperties service at all?
Conclusions:
-
we have an official naming convention
-
need to be sure ArgoNeuT is still supported
-
Gianluca Petrillo and Jonathan Paley will talk for the hand off
-
the possible
LArProperties simplification will be delayed
There are minutes attached to this event.
Show them.