Software phone meeting on 12 January 2006

Present: Bruce, Eric, Florian, Gilles, Kambiz, Murrough, Norman, Paolo, Rainer, Richard, Victor.

Status reports

Birmingham: experience with PMP Presenter

Gilles has been trying the PMP Presenter to view histograms. He has a few problems with setting up the configuration and the display options. He will summarise these to Paolo for advice and/or fixes.

Heidelberg: TileCal test SW and PPM monitoring

Florian has made some code cleanups and added overview files with all histograms also from pedestal runs. His ability to test new code in Heidelberg is limited as there is no longer a TCM there.

Kambiz reported that strange behaviour (truncation?) has been seen in the VME readout when reading large (eg 100) raw slices from the PPM. This needs to be investigated.

Victor has started work on a new ppmMonitoring package. He has started by creaqting ROOT trees from PPM data. The next steps are to provide bytestream converters from the VME readout (as agreed at the previous meeting).

Florian asked how the multistep run step number will be tracked when using ROD readout. There is a bit field within the detector event type that we have foreseen for this purpose.

Mainz: JEM production testing SW

Rainer has started to work on the production test software for the JEM. He will be at CERN for the whole of the week when we have our software meeting so some further detail discussions could take place then.

QMUL: ROD crate DAQ

Murrough reported that readout of our modules into the monitoring stream was now working under the ROD crate DAQ controller. More work is still needed to complete and clean up the code, but now all the basics for moving our run controllers to ROD crate DAQ are in place.

Bruce remarked that we still need to decide how VME readout from non-ROD modules will be flagged in these events which have ROD fragment style headers.

RAL: connectivity, COOL, SimpleBuffer

Norman has made some updates to the new connectivity package.

Richard has been working on storing his analysis data in ROOT trees. He and Norman are just starting on this work and would like to follow Florians examples if possible. Florian commented that for custom ROOT classes we need a way to create the ROOT dictionary classes. This may already be implemented in Adrians histoDisplay package -- to be checked.

Richard has has also developed his CMM calibration procedures. To test this he would like more input modules. A test with many CPM should be arranged at Birmingham.

He has also looked into accessing the internals of COOL tables using the lower level RAL software. This may help in selecting which queries to make via the COOL API.

Norman reported continuing issues with the multistep run procedure, in particular unreliability at the first and last steps and that the number of steps is also not always reliable.

Bruce has prepared two sets of changes to the SimpleBuffer interface in the infraL1Calo package. One changes the API to look more like that of std::vector. The other will allow module readout to write data as structures rather than just successive push_back() calls. It was agreed that he should now commit these changes (both in infraL1Calo and the affected packages), tagging between each of the two changes.

Florian emphasised that he does not want to change the way the present PPM readout works. It will not fit neatly into fixed structures as it is inherently variable length.

Feedback to GNAM and Presenter Meeting

Paolo will attend the forthcoming meeting of GNAM and Presenter developers. He asked for our feedback to the meeting. Unfortunately so far we have not used GNAM in earnest so we dont have any feedback yet. He will forward the problems reported by Gilles with using the Presenter. Kambiz asked about documentation. It seems the GNAM documentation was lost in a machine crash. So first among our feedback is the request for new documentation to be prepared. Until then, there are some papers describing the packages. Paolo will circulate a list of references.

Switching between Debug and Optimised Versions

Norman has asked how to switch between using the debug and optimised versions of code. In principle changes are needed in two places. One is the .l1calo_bashrc script to set the CMTCONFIG variable used for building L1Calo software and when running interactive programs. The other is in the Partition database object for the partition you are using where you need to set the "DefaultTag" to be used.

January Software Workshop

We agreed to move the dates for this workshop to 30-31 January. The exact agenda is still to be defined.

It was confirmed after our meeting that this week is indeed a LAr online software experts week and that we should be able to arrange to talk with them.

Run Control States

After our last meeting, Giovanna kindly put the proposed changes on hold until we could agree our response.

In emails after that meeting, we felt that provided we could, if necessary, still use synchronised load and configure steps as internal substates, that we could agree to the change in the global states. Murrough should double check this with Giovanna and if this is OK, convey our agreement.

After the meeting, I checked this with Giovanna. She confirmed that it would be possible to give our run controllers separate rules so that, as part of the new single Configure transition, we could execute a synchronised action (eg our present load method)in ROD Crate DAQ before the configure action was executed. She will prepare a little demonstration and if this is OK I said she could go ahead with the changes which should then make it into the next TDAQ release as planned.

As part of the discussion at this meeting, Norman raised the issue of how the calorimeters will use the transition states between steps in multistep runs, ie how much they need to reconfigure to change energies between steps.

Action on Norman to arrange a discussion with both calorimeter groups.

Production vs Development Software

Bruce asked how we ensure we can easily develop software while also being able to have frozen production releases. Florian also asked in particular about development of software for use at point 1.

The present scripts for production releases allow two cases. Firstly, an official release of packages with selected tags (a list prepared by hand after one or two rounds of emails). Secondly, mainly intended for point 1, a release based on the latest tagged versions of each package, with the list of latest tags prepared and saved automatically. This is intended for use with the l1calo account for production testing.

In addition, it is possible just to copy a personal development release (based on HEAD versions from CVS) to individual user accounts at point 1. This can be used for testing.

The danger is that these can diverge if packages with new code are not tagged. So we do now require that new developments are tagged after being tested so that production (tagged) releases and development (HEAD) versions are not too far out of step.

Infrastructure in 3150

Kambiz asked about the existing computing infrastructure in our 3150 office. At present we have two working computers and two screens. One screen on order has not yet been delivered and one of the new PCs has an intermittent fault with its ethernet card and probably needs to be sent back for repair.

On a related topic, there is now a spec for the rack mounted DCS PCs soon to be ordered by Helfried Burkhardt. Murrough will circulate this as we may want to buy the same for our racks. At least we should ask for one from the first batch for test purposes.

It was also asked when some of the cabling benches might be moved out of the 3150 lab space. In fact some were moved out during the meeting though its not know if this is permanent or not.

Next meetings

The next phone meeting will be on Thursday 26 January at 11:00 CET (10:00 GMT). The software workshop will be held at CERN from 30-31 January.