Librarian's Meeting (06 May 2014)
Chaired by: Snider, Erica
Partecipants:
B. Baller
L. Garren
H. Greenlee
T. Junk
G. Petrillo
B. Rebel
W. Seligman
E. Sexton-Kennedy
E. Snider
T. Usher
Release polices
- a summary of the policy is given and received
- the major version of LArSoft should allow for two digits
Status of releases
- clarified the possibility of having multiple tags for the same release, one
for each experiment needing any
- policy for deletion of releases described and received
- question about the relation between release tags (in GIT) and UPS versions
- v1_00_00 is waiting for validation from the experiments:
* MicroBooNE (H. Greenlee) has found a bug, probably a pre-existing bug which
the updated ROOT version now exposes; will be fixed in a patch release
* LBNE (T. Junk) has not proceeded to validation, but asks us not to be
stopped by that
We ca now proceed to merge v1_00_00 to the develop branch
Test/Validation workshop:
- presented the scope of the workshop, promised more details within this week
Random Number Generation:
- B. Rebel asked to be warned before the change in Genie random numbers
happens
- clarified that the changes in the Genie interface to LArSoft, in nutools and
shared by other experiments, is in the hands of nutools (R. Hatcher)
ROOT output size:
- presented figures from some 100-event test cases
- an objection was raised that the performance will be much different in
a test with larger event pool; such tests will be performed
- performances might be sensibly different with different compression
algorithms ROOT provides (we use zlib as it is default)
- the uncompressed output of LBNE (full detector) simulation raises some
questions; tetss will repeated with the Region of Interest coded in the wires
- it is up to the experiments and analysers to decide which compression level
to use, possibly different for different jobs
Other topics:
- mrb v1_00_00:
* L. Garren asks about early testers of v1_00_00, not yet made default
* since no feedback is given, the version will be pushed as default as soon
as a couple of known issues are addressed
* W. Seligman reports problems with UPS setup in zsh
- B. Baller advertises documentation on his recent changes to the
ClusterCrawler algorithm, raising the question where it should reside;
proposals from the partecipants include:
* a (new) LArSoft DocDB: drawback is that it must be (manually) linked into
experiments' DocDB too in order to make search easy
* as a talk in Indico: not handy to search for
* in redmine: also not that handy to search for
The proposal by E. Snider of a LArSoft DocDB seems to be in principle well
accepted, but no action is decided.
There are minutes attached to this event.
Show them.