CORAL Coffee

Monday 11 Aug 2008, 10:00 → 11:00 Europe/Zurich

Resto 2 caffeteria (CERN)

Zsolt Molnar Molnar

Informal meeting about CORAL issues next to a coffee. CORAL developers meet with the representatives of the persistency projects that depend on CORAL (mainly POOL and COOL). Suggested topics: - Overview of CORAL bugs and requests status (bug asignee-s) - CORAL release plan. Confusing message propagated down from AA meeting, should be resolved (Andrea) - CORAL activity plan and task distribution (Zsolt). Tasks: - Bugfixes - OSX client integration - fixing failed tests - fixing the nightlies

Participants: Andrea Valassi, Zsolt Molnar CORAL tests • Alexander finishes the backport of CORAL 2.0.0 tests to the CORAL 1.9 branch. Action: Alexander • The nightlies should execute the CORAL unit tests as well. Start thinking about it. Follow-up: Friday (15 Aug). Action: Alexander. CORAL bugs: • Fix bug #38968: MySQLAccess problems in readback of unsigned char. Action: Zsolt CORAL extensions: • We should think about the URL format: Reiner's suggestion or the existing one. But this case is not urgent. Collect and provide information about it. Action: Andrea. OSX Oracle client integration • Experiments requested it explicitly, so we will start it this week, following Dirk's recommendations. Finish it by the end of the next week. Action: Zsolt CORAL nightlies status • CORAL build problems (compilation and missing components) fixed. Update functionality • We start it soon, after the design document. We are behind the schedule with it. Design document • We provide a skeleton of the design, not a fully-blown one. We do not have templates, as none of the persistency projects provided such a document. The CORAL server project will be the first one. Action: Zsolt (overview, Coral{Access,Server,Protocol}), Alexander: CoralAccess. CORAL server profiling • Finish the profiling, tbd in the CORAL server meeting. Action: Zsolt. Andrea's coments * CORAL release plan. The Coral server/client functionalities will only be released in CORAL210. This is consistent with the original plan circulated by Zsolt/Dirk to the AF (first release end October 2008). The ongoing work on the SEALed 19x branch is meant for Atlas tests only but is not going to be released as a production version for data taking with the first LHC beams. The next CORAL210 release may contain only the read-only coral server/client functionality if DML/DDL are not ready yet. Unless urgent bug fixes are requested, no CORAL 198 or 201 releases are planned with binary compatibility for 197 or 200 respectively. There is no confusion left about the CORAL release plan at the moment. * Windows status and plans. CORAL dev1 build is successful. This indicates there are no problems with unresolved or duplicate symbols in this slot (bug #35407). CORAL dev build fails on Boost headers. Once this is solved, linking problems may appear (bug #35407). Need to contact Stefan about Boost. Action: Andrea (done). CORAL build warnings on Windows (see dev1) should be removed. We should start the port to VC9 according to Stefan's instructions. * POOL nightlies Problem from missing PyCoral from CORAL is now fixed. One pending failure. Action: Andrea contact Witek (done). * Exception handling (bug #39945) This is not a bug but rather a task needed to support the CORAL server. New exceptions will not be released until CORAL210. Will rediscuss details before the release. Action: Andrea (make comments if any). * CoralBricks Pragma header for boost 1.35 warnings is temporary (should be removed after the migration to boost 1.36). CoralBricks contains several headers that are needed in private impementation code but not in any public API headers. For boost threads this defines a typedef that could allow easy switching between debug/nondebug boost threads inside the code. Another header deals with boost xtime. Some of the new headers could be in CoralKernel, but CoralKernel does not exist in CORAL19x. As the next new release will be CORAL210, should rediscuss whether this new package is really needed or headers could be added (for instance to CoralKernel). Action: Andrea (make comments if any). * CORAL server URL format If we eventually choose Reiner's URL proposal, COOL will not work unless the CORAL API provides a hook to tell users what is the technology of the remote database (sr #103545). * Design document. Andrea would like to see one document describing how the client and server are integrated at the software level (rather than two separate documents). Andrea would like to have this document before the implementation of DML and DDL functionality is started, and possibly by the end of this week. The initial draft of this document can be quite simple. The style could (but is not required to) follow the format used for Andrea's design document https://espace.cern.ch/project-coralserver/Documents/Design%20proposal%20for%20client%20and%20server/CoralClientServerDesign.docx (We did not discuss that this design document should/would become the template for any further documents provided in the PF).