GR v15r33 (Heather) We just tagged GR v15r33 in preparation for the move to fsw B1-1-0. On Thursday, July 31st, we anticipate an official move to this version of fsw on board. Anders and company are preparing to use GR v15r33 in the L1proc by that time. This tag of GR includes all the upgrades necessary for the move - OnboardFilter, updates to digiRootData and the ldReader. [Meanwhile Eric C. and Joanne have made changes to be compatible with fsw B1-1-0 to non-GR code invoked by Online — packages fswDecipher and MOOT.]
New fsw release (Heather)
The major change is the appearance of a new version of the gamma filter, where the interpretation of the status bits now varies depending on whether it is V0, V1, or V2. Due to the limited amount of time to prepare, we had a meeting last week and decided that we would make the status bits available (as we already do) and in addition provide the version so that users can unpack the bits on their own. To this end, merit has a new column, FswGamVersion. In the future, we anticipate providing access to the 32 new accessor methods that are actually available in this new version of fsw.And from Tracy:
We completed the upgrade to B1-1-0, including verifying against some test evt files Bryson made from ebf data files made by Gregg on the test bed. Gregg made those files running the gamma_normal_leak configuration, so there were differences in those events (because Gleam doesn't see all the events the test bed sees, it can't leak at the "right time"), otherwise there was agreement in all but 1 of some 120,000 events. The one event is under investigation but is certainly not a show stopper.One thing discovered was that a fix to B1-1-0 which appears to have not been implemented was to fix the "negative energy overwrite" problem, where a negative energy, packed into the lower 24 bits of the stage/energy word, overwrites the stage bits (the upper 8 bits of the stage/energy word). Next release will have that.
Trigger (Heather) Zach is testing out the elimination of GltDigi. Johann is going to change TriRowBits from an algorithm to a tool.
G4 propagator (Tracy)
Johann's latest test jobs still have some failures with the same symptom, where a particle gets "stuck" and begins an infinite loop of being stuck and spitting out messages, eventually crashing when the log file fills. This is very suspicious as the G4Propagator package he was using was specifically set up to not only prevent stuck particles, but if it did somehow happen to get stuck, to limit the number of messages before throwing an exception and moving on. Since it didn't do that it makes me think we are executing another section of code which I had thought could not ever be executed. Well, I guess if code is there it can execute! So, am trying to fix that piece of code and try again.
Meanwhile Leon has improved handling downstream:
When TkrReconAlg throws an exception (so far, due to G4Propagator getting bad input or getting stuck), it will clean up by erasing all tracks and vertices, and unmarking all used clusters. A bit drastic, but it only happens once in a blue moon, and it's not obvious what would be better. The event then becomes "CAL-only"
Alignment (Toby) All the relevant code [his and calibration infrastructure it uses] is in CVS and tagged, except for a change to job options. He ran into an early problem checking it out on Windows which [Joanne hopes] may be resolved by making a clean build of everything, but he has not had a chance to follow up. Most likely the code in CVS is ok or very close to ok. [Update: there was a bug in CalibSvc — failure to declare new LATAlignment converter - which caused problems only on Windows. That is now fixed in a new tag, v0r39p2]
Reprocessing (Heather) Tom is EVO-mute this morning, but reports "Work steadily ongoing. Progress. Small tests already started. Real soon now..."
There are two new files available; a Data Access Help file on the LAT Data Processing Monitor tool and, thanks to Jim, a new Science Tools Reference/Parameters file on gtexposure.I'll also be making a new tag (v6r0) for the WB, so Berrie can update the mirror site.
Skimmer: (Heather, David C.) Summary from Heather:
Warren has updated the next version of the L1 processing merge script to rebuild the indices for the full ROOT trees: digi, recon, relations, gcr. This will help mitigate the ongoing issues with skimming some of these files. We have already updated RootIo to rebuild the indices from scratch, so currently there should be no troubles with viewing events randomly in the event displays - assuming the file itself is well formed.There are some ongoing issues concerning the skimmer: Tony has been studying reports of trouble using the web interface for the skimmer (5.0) on merit files. Tony has tracked this down to one specific merit file. The jobs seem to fail with a large use of memory. I can run a TCut in ROOT interactively on this file and it works out just fine, with little memory usage. The file in question does not have a tremendous number of events, so it is unclear why it would be causing trouble.
There are other reports of problems with the version 6 series of the skimmer, but we didn't get into that at the Core meeting today. But we are happy to see David is available to assist.
After sending a reminder about packages that still needed to be tagged to include the changes to environment variables a lot of packages were tagged. As far as I can tell, these packages all combined fine in the RM. I did see 3 unit test failures in LATEST builds of the RM but they don't seem to be related to my changes. I'm just about ready to have the new RM tag GR packages with LATEST tags for SCons to use. I sent out an email about this yesterday and if I don't hear any complaints I plan on turning this on today or tomorrow. I have several feature requests pending from Joanne to include in the RM libraries so that they can be used in GoGui. Once I have the RM tagging LATEST packages I'll start working on those.Something that I forgot to mention during the meeting is that I had sent a request on how to deal with the GR job options files. I had not yet received any input on my suggested approach. I will be resending that request to see if there's any feedback or if I can go ahead with the proposed approach.
(Emmanuel) is about half-way through making SConscript file for all the GR packages. ETA is Friday of this week or Monday of next.
(Joanne) GoGui has made some progress since last week. A package within a full installation like ScienceTools can be selected and relevant files displayed. It's also possible to change from one root installation directory to another. Next areas of concentration will be more work on saving/restoring options and, as Navid provides service routines, ability to do CVS and SCons operations.
(Heather) Richard is contemplating another Core meeting around Halloween.
(Leon)A problem with the EventAuditor is that it has been messing up the end-of-run Chrono table with extraneous and incorrect entries (although the correct entries are still there). I think this is one of the reasons that it's remained unofficial. I've fixed that problem at the head of AnalysisNtuple. This won't appear in GR until after the FSW mods settle down.
previous | minutes index | next |