Date
Jan 04, 2017
Attendees
John Clyne, Scott, Alan, Sam
Goals
Agenda
- Admin updates
- VAPOR Build issues
- ASD
- round table / progress updates
Notes
Problem with VAPOR_HOME in win10 install occurs when software installed for personal use (not system use), and is not unique to VS2015 (occurs with VS2010 too). A sev 6 bug has been reported and there are options for fixing or workaround around this problem
GetAppPath class is only partially implemented for windows. Does not support use when running from a source build. When running from source on windows setting VAPOR_HOME is required
Discussed python 2.11.12 bug that impacts header include order on Mac systems. This is a known python bug that has been fixed in python 2.11.13. It does not look feasible at this time to support 2.11.13 on windows due to use of Anaconda build. Two options: 1. Keep python builds consistent across all platforms at 2.11.12 for now and deal with the bug the next time we encounter it (our code has been changed to work around the problem), or 2. Use 2.11.13 on Mac (and possibly Linux) only. We seem to have defaulted to option 1.
Sam has the apps-2017 tree built for Mac. However, three of libraries use different versions than the windows apps-2017 tree. These are Qt, geotifff, and something else. We need to try to resolve this difference. Sam built all python modules from source. Discussed using PIP to provide external python modules (e.g. Numpy, scipy, seedme, etc). Also discussed using Anaconda, which supports all three platforms.
Alan has not yet merged the changes needed for the VS2015 build into the main git branch. This can be done as soon as Scott and Alan are ready.
Sam suggested we migrate from sourceforge to github. John wants to hold off until after the release of 2.6.0
Sam also suggested using cmake for building vapor. Again, John suggested tabling discuss until after 2.6.0 release and possibly limiting new build system to vapor3
Sam indicated that there are still problems with getting a usable parallel version of visit for Dimitri's ASD work. Dimitri's team wants to do in situ visualization, and thus need a high scalability with VisIt library (so far only scaling to 32 nodes (cores?). John is concerned that even if technical problems with VisIt are resolved performing in situ vis with Cheyenne scheduling system, etc. may be problematic. SSG should probably be brought into conversation.
Action Items
| Item | Who |
|---|---|
Explore using Anaconda for python builds on all three platforms (need to verify ok for Mac and Linux) | Sam |
| Set up wiki page with table of all library versions used for apps-2017 tree | Alan and Sam |
| Sync up 3rd party lib versions for apps-2017 tree on Mac and Windows | Alan and Sam |
| Set up meeting to better understand Dmitri's needs and make sure they can be met | John |