Re: [petsc-dev] [JIRA] (SDK-82) Complete a manual test run of xSDK on Edison

2016-04-19 Thread Barry Smith
Jason and Alicia I have attempted to fix the --with-batch option to work properly with the various --download-xxx Could you please try your installs again with this option? Send configure.log on failure to petsc-ma...@mcs.anl.gov Thanks Barry Technically we should always

Re: [petsc-dev] Fwd: [JIRA] (SDK-82) Complete a manual test run of xSDK on Edison

2016-04-19 Thread Jason Sarich
> >So on this machine you can compile and run programs on the front end > but if they call MPI_Init() then you cannot run them? mira has this same situation. Does it make sense to add a command line option to configure to prevent the running of mpi-enabled executables (or perhaps all third

Re: [petsc-dev] Fwd: [JIRA] (SDK-82) Complete a manual test run of xSDK on Edison

2016-04-19 Thread Satish Balay
Hm - PETSc configure does not need to run MPI jobs with MPI_Init. So it uses this quirk to get-by without requiring --with-batch And it also tries not to run testsuites from externalpackages. [clearly some packages are not doing this] Perhaps there should run a test with and without MPI_Init()-

[petsc-dev] Fwd: [JIRA] (SDK-82) Complete a manual test run of xSDK on Edison

2016-04-19 Thread Barry Smith
Satish, So on this machine you can compile and run programs on the front end but if they call MPI_Init() then you cannot run them? What does this mean with regard to --with-batch? It seems we don't need --with-batch on that machine yet it cannot run MPI programs? Barry > Begin